A Qualitative Review of Full Sized Autonomous Racing Vehicle Sensors:

A Case Study

Manuel Mar

and Eric Dietz

Department of Computer Information Technology, Purdue University,401 Gran St, West Lafayette, U.S.A.

Keywords:

Autonomous Vehicles, Racing, Motorsports, High-Speed, Sensors, Autonomy Hardware,

Vehicle Architecture, High-Performance, ADAS, Electric Vehicles.

Abstract:

This paper explores into the challenges and advancements encountered in the development and operation of

full-sized autonomous cars built for motorsports competitions. Concentrating on a qualitative examination

of the sensor conﬁguration, structure, and real-time assessment of vehicle platforms in the Indy Autonomous

Challenge and Roborace. The scrutiny is centered on recent years’ research and the vehicles’ performance

in demanding conditions, systematically highlighted and summarized in this paper. The analysis furnishes a

more concise and condensed comprehension of the prevailing trends in such competitions, offering insights

into the future of autonomy in the coming years.

1 INTRODUCTION

Advanced Driver Assistance Systems (ADAS), as de-

ﬁned by the SAE, encompass six levels of automation

(On-Road Automated Driving (ORAD) Committee,

2021). These systems heavily depend on an array of

sensors and software to accurately perceive their sur-

roundings, to achieve full automation without human

intervention. As we move closer to this reality, the

sensor industry is witnessing rapid growth, innovat-

ing to meet the challenges that autonomous systems

present (Ahangar et al., 2021). Notably, autonomous

ground systems such as cars, trucks, and trains con-

tinue to grapple with speciﬁc, unresolved challenges,

motivating researchers and engineers to dive into this

area (Yeong et al., 2021).

The world of motorsport, characterized by condi-

tions like steep inclines, high-speed cornering, and the

nuanced techniques such as”lift and coast”, presents

its unique set of challenges. Racing circuits featur-

ing vehicles like IndyCar, Formula E, and Formula

1 represent the pinnacle of high-performance design.

The technological innovations nurtured in these rac-

ing arenas often ﬁnd their way into commercial vehi-

cles (Sarkar and Mohan, 2019). Racing drivers, with

their deep understanding of vehicle dynamics and

performance, exhibit skills and techniques that are

difﬁcult to replicate via software or automated sys-

https://orcid.org/0000-0003-3245-7964

tems. While sensors can process information faster

than human senses, the nuanced comprehension a

racer possesses often surpasses that of an average

driver. In recent times, numerous autonomous rac-

ing competitions have emerged that challenge engi-

neers and researchers (Buehler et al., 2009),(Robo-

race, 2016),(A2RL, 2023),(IAC, 2020). These pro-

grams provide a platform to gain deeper insights, ad-

dress existing issues, and elevate vehicle performance

and development. Some of these competition mile-

stones are shown in Figure 1.

2008 DARPA Challenge

2016

Roborace - Devbot 1.0 – Initial Deployment

2017 Roborace - Robocar Testing

2019 Roborace - Robocar - Speed Record - 284 kph

2019 Roborace - Devbot 2.0 – Multi-vehicle Competittion

2020 Roborace - Devbot 2.0 – Collision Avoidance Test

2021 IAC - Single Vehicle Trials Race

2022 IAC - Multiagent Race at LVMS

2023 IAC - Road Course Single Vehicle Time Trial

2024 A2RL - Multiagent Abu Dhabi YAS Marina Circuit

Figure 1: Autonomous Racing Milestone.

High-speed racing vehicles, as deﬁned within the

context of this research, encompass vehicles engi-

neered to operate under stringent conditions, expe-

riencing substantial lateral forces, and capable of

achieving swift acceleration. Over the past decade,

technological advancements in sensors have spawned

Mar, M. and Dietz, E.

A Qualitative Review of Full Sized Autonomous Racing Vehicle Sensors: A Case Study.

DOI: 10.5220/0012634800003702

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 10th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2024), pages 311-318

ISBN: 978-989-758-703-0; ISSN: 2184-495X

311

an array of high-performance ground vehicle pro-

grams, each characterized by distinct metrics, propul-

sion systems, and performance outcomes. The last

decade has introduced multiple autonomous racing

programs where ground race vehicles were built

speciﬁcally for racing purposes, vehicles are han-

dled at their limits (Betz et al., 2022). This research

will concentrate on two signiﬁcant projects that in-

volved the development of full-sized autonomous rac-

ing vehicles for racing platforms: Roborace and the

Indy Autonomous Challenge. These programs show

distinctions in their racing architecture, conceptual

frameworks, and performance metrics. Nonetheless,

they also share commonalities, particularly concern-

ing sensor and software performance in real-time sce-

narios. Building upon the foundation laid by (Mar

et al., 2024), this study will predominantly delve into

the sensor setup of each vehicle model, describing

speciﬁc challenges these sensors may encounter dur-

ing testing. Subsequently, we will offer a comprehen-

sive overview of their real-time performance based on

previous tests.

2 SENSOR SETUP

When considering the vehicle hardware conﬁgura-

tion, the selection of speciﬁc devices involves several

possible combinations inﬂuenced by both technical

and non-technical factors. Similar to traditional mo-

torsports competitions, these selections are not only

dictated by technical considerations but also by non-

technical factors, notably the availability of sponsors.

It’s noteworthy that, unlike conventional racing, spon-

sors in autonomous racing have a direct impact on the

ﬁnal vehicle design and the choice of hardware com-

ponents.

In this section, we present a comprehensive table

summarizing the sensors used in the vehicles from the

studied competitions, namely Roborace and the Indy

Autonomous Challenge (IAC).

2.1 Roborace

Roborace, inaugurated in 2016, introduced three dis-

tinct vehicle models: Devbot (2016), Robocar (2017),

and Devbot 2.0 (2018). While there are notable dif-

ferences in design across these models, the sensor ar-

chitecture remained consistent for the three different

versions of this series. The sensor suite included Li-

DARs, cameras, radars, and ultrasonic speed sensors

as the extereoceptive sensors, which are shown in Ta-

ble 1. The computational backbone consisted of the

NVIDIA Drive PX2 (NVIDIA, 2016) for high-level

planning and perception processing while the Speed-

goat Mobile Target Machine (Speedgoat, 2016) was

used for real-time control tasks (Betz et al., 2019) and

ensuring low-latency communication with the vehi-

cle’s actuators, it facilitated rapid adjustments based

on the decisions made by the high-level planning sys-

tem.

The integration of LiDARs, cameras, and radars

systems provided a 360-degree environmental percep-

tion. Four LiDARs working in tandem with six cam-

eras contributed visual input for object recognition

and enhanced understanding of the race environment.

Additionally, four radars added layers of sensing, re-

inforcing the vehicles’ ability to detect and respond to

surrounding objects. This sensor fusion strategy cre-

ated a sophisticated perception framework, boosting

the vehicles’ overall awareness on the race track.

The transition from Devbot 1.0 to Robocar and

Devbot 2.0 exempliﬁed a systematic approach to sen-

sor placement and testing objectives. The initial de-

sign of Devbot 1.0/2.0, featuring a Le Mans Proto-

type (LMP) chassis, carefully considered scenarios

involving potential human intervention. In contrast,

the cockpitless design of the Robocar gave a distinct

futuristic design, stressing the intent to test the vehicle

at elevated speeds without direct human intervention.

This design choice reﬂected a forward-looking vision,

emphasizing a commitment to pushing the boundaries

of performance. Furthermore, the powertrain design

displayed a deliberate effort to replicate the advance-

ments achieved in the electric Formula E, leveraging

its widespread popularity. This emphasis is notably

reﬂected in the choice of electric motors as the pri-

mary propulsion source for these vehicles, with de-

tailed power speciﬁcations provided in Table 1.

2.2 Indy Autonomous Challenge

Introduced in 2020 and hosting its inaugural race in

2021, this series debuted with the AV-21 model, ini-

tially engineered by Clemson University’s Deep Or-

ange project (Zhu et al., 2021). Mimicking the phys-

ical appearance of the Indy Car series and adopting a

cockpitless design, which emphasizes the avoidance

of direct human intervention during testing.

The selection of sensors, detailed in 1, aimed at

redundancy for decision-making and perception. Li-

DAR, Radar, and Cameras were chosen, offering ﬂex-

ibility for usage individually or in fusion. This selec-

tion is intended to equip the vehicle with a compre-

hensive understanding of its dynamic environment,

crucial for navigating at high speeds. LiDAR tech-

nology, with three units, played an important role in

addressing challenges such as reﬂection delays and

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

312

Table 1: Sensor Summary Roborace and IAC Vehicle Model.

Sensor Type Device Made Model Series Vehicle Model Quant

Perception

LiDAR

Luminar Hydra 3

IAC

AV-21 3

Luminar Iris AV-24 3

Ouster Ibeo - OS1 -16/64 Roborace D1/D2/Robocar 4

Radar

Aptiv ESR 2.5 - MRR

IAC

AV-21 2

Continental ARS548 RDI AV-24 2

N/A N/A Roborace D1/D2/Robocar 4

Camera

AlliedVision Mako G319C IAC AV-21 6

N/A N/A Roborace D1/D2/Robocar 6

Localization GNSS

Novatel PwrPak 7

IAC

AV-21 2

Vectornav VN-310 AV-21 /AV-24 1/ 4

N/A N/A Roborace D1/D2/Robocar 1

Powertrain

Engine

N/A Electric -136kW Roborace D1/D2/Robocar 4

Honda Honda K20C IAC AV-21/AV-24 1

ECU

Motec M142

IAC

AV-21 1

New Eagle GCM 196 Raptor AV-21 1

Mclaren N/A Roborace D1/D2/Robocar 1

Communications Switch

Cisco IE 3300

IAC

AV-24 1

Cisco IE 5000 AV-21 1

Computing

CPU

Dspace Autera Autobox

IAC

AV-21/AV-24 1

Adlink AVA 3501 AV-21 1

Speedgoat MRT Targetmachine Roborace D1/D2/Robocar 1

GPU

NVIDIA NVIDIA Drive PX2 Roborace D1/D2/Robocar 1

NVIDIA Quadro RTX 8000

IAC

AV-21 1

NVIDIA RTX A5000 AV-21/AV-24 1

adapting to banking angles. The Radar introduced

an additional layer of redundancy, enhancing the ve-

hicle’s capacity to detect and respond to dynamic

changes in the racing environment for medium and

large-range detection. The multi-camera setup fa-

cilitated comprehensive visual coverage, contributing

to object recognition, lane tracking, and an overall

understanding of the racing environment (Ayala and

Mohd, 2021). While originally designed for indus-

trial applications such as surveillance, machine vi-

sion or robotics,the cameras were repurposed for au-

tonomous racing. Which were not intended for expo-

sure to high lateral forces and vibrations. For local-

ization, two GNSS units were used initially; however,

due to some challenges during initial tests an addi-

tional unit was inserted later which was intended to

address signal loss or inaccuracies at higher speeds.

All sensors had the highest refresh rates in the mar-

ket, a crucial feature for a racing context, where split-

second decisions are imperative. IAC AV-21 incor-

porated two robust embedded computers equipped

with the NVIDIA Quadro RTX 8000 GPU and the

RTX A5000. Additionally, the selection included the

ADlink AVA-3501 and Autera Autobox. These de-

vices were selected mainly because of the harsh con-

ditions this car experimented with, embedded sys-

tems by deﬁnition are designed to perform speciﬁc

tasks (Tumeo et al., 2017). The powertrain conﬁgu-

ration remained consistent with that of an Indy Car,

which was not altered. Additionally, the choice of

data transfer mechanisms and electronic control units

(ECU) was provided by combining industry and com-

mercial devices, resulting in data acquisition rates that

exceeded standard levels.

3 REAL TIME PERFORMANCE

The real-time performance evaluation of autonomous

racing vehicles is crucial to assess their capabili-

ties and address challenges encountered during high-

speed racing scenarios. This section delves into the

dynamic aspects of the Roborace and IAC vehicles,

emphasizing real-time challenges and outcomes.

3.1 Roborace

As this remains a motorsports competition, the speeds

achieved during each event or race are of signiﬁcant

importance. The summarized speeds, as shown in Ta-

ble 2 exhibit a gradual acceleration in both Roborace

and the IAC. At the same time, competitors require

extensive software development and validation, a sub-

stantial portion of which occurs off-the-track.

A Qualitative Review of Full Sized Autonomous Racing Vehicle Sensors: A Case Study

313

3.1.1 Devbot 1.0

The preliminary rollout of the vehicle underwent test-

ing at UK Donington Park (BBC, 2016) and the Mar-

rakech Formula E Street Track. During the latter ses-

sion, the vehicle completed 12 laps in a time frame of

30 minutes (Knight and Blendis, 2016). These tests

were conducted to assess and evaluate the initial de-

ployment of sensors and the overall real-time perfor-

mance on the track. After that, the deployment of De-

vbot 1.0 at the Buenos Aires EV Grand Prix in 2017

witnessed two signiﬁcant events. In the ﬁrst event, the

vehicle completed the course track, reaching a maxi-

mum speed of 186 km/h. In the second event, a colli-

sion occurred when one of the vehicles miscalculated

a corner while traveling at high speed, resulting in

a crash (Kelion, 2017). ’Devbot’ raced against hu-

mans in two different experiments. The ﬁrst one on

the Hong Kong Central ePrix track, in which a non-

professional driver got 86 seconds compared to De-

vbot’s time of 94 seconds, surpassing Devbot by 8

seconds, where both reached top speeds of 150-160

km/h (Dow, 2017). In 2018, a professional driver re-

peated the experiment and raced against Devbot 1.0

in Rome for the opening of the Formula E event. In

this case, the pro-driver outperformed the Devbot by

26 seconds(Fingas, 2018).

University teams initiated the testing of their soft-

ware platforms by executing three autonomous laps

on the Berlin Racetrack, achieving top speeds of up

to 150 km/h. The use of a global optimal planner for

path generation ensured smooth on-track performance

with no notable hardware issues or delays. The lateral

error was minimal, reaching a maximum of 0.8 me-

ters (Stahl et al., 2019b). Another study conducted by

(Caporale et al., 2018) demonstrated a low lateral er-

ror of 0.3 meters. This study employed sensor fusion

state estimation and a non-linear MPC controller in

the initial version of Devbot on a road course track for

two laps. In both instances, the computing and sen-

sor architecture exhibited robust performance with-

out signiﬁcant issues, allowing the vehicles to achieve

high speeds of up to 150 km/h.

In (Caporale et al., 2019), various challenges were

identiﬁed in the autonomous vehicle system. One no-

table issue pertained to computing, speciﬁcally the

overload of the ARM CPU during scan matching on

the PX2, resulting in occasional failures. Addition-

ally, concerns were raised about vehicle alignment

during trajectory planning. The use of a single mass

model failed to account for instances where the vehi-

cle’s alignment did not align perfectly with the path

tangent, consequently leading to reduced accelera-

tion, particularly in certain turns.

A ﬁnal test was performed with the Robocar

where the vehicle was pushed to the limit and reached

280 km/h to set a new record (Roborace, 2019), there

has not been any disclosure of data or study that was

done during this test.

3.1.2 Devbot 2.0

Devbot 2.0 experienced extensive testing across var-

ious race events, including the Zala Zone, Circuit de

Croix, Montebanco Spain, Modena, and others. In a

study conducted by (Stahl et al., 2019a), the vehicle

planner achieved an average rate of 16.8 Hz, demon-

strating capabilities up to 212 km/h with a 200ms

prediction to anticipate the movements of the lead-

ing vehicle. Tests carried out at the Zala Zone Hun-

gary and the Circuit de Croix-en-Ternois in France

showcased lidar-based localization with a lateral er-

ror consistently below 10 cm. The vehicle achieved

speeds exceeding 45 m/s and accelerations greater

than 10m/s

. Challenges arose in time synchroniza-

tion with sensors, as a 10ms delay at 30m/s could

result in a 0.3m error, necessitating vehicle odome-

try (Schratter et al., 2021). (Renzler et al., 2020) ad-

dressed lidar distortion correction and delay compen-

sation at Zala, reaching speeds up to 90 km/h with

accelerations of 10ms at a 20Hz rate . Teams also

experimented with Kalman ﬁlters in conjunction with

LiDAR, IMU, and vehicle dynamic sensors, achiev-

ing a peak speed of 90 km/h. (Zubaca et al., 2020) set

a lap record of 1 minute and 37.440 seconds at Cir-

cuit de Croix-en-Ternois, averaging approximately 65

km/h.

Some of the performance and metrics parameters

are:

• State Estimation. The Improved H-inﬁnity Fil-

ter, utilizing vehicle sensors like LiDAR, IMU,

GPS, and Vehicle Odometry, consistently main-

tained estimation errors across laps. In con-

trast, the Extended Kalman Filter (EKF) exhib-

ited growing errors after each lap (Zubaca et al.,

2020).

• High-Speed LiDAR Use. The study demon-

strated that even when subjected to high speeds

and accelerations reaching up to 10 m/s² in both

longitudinal and lateral directions, precise LiDAR

measurements and corrections can be achieved

(Renzler et al., 2020).

• Effect of Distortion on Dynamic Driving. Dis-

tortion was less visible when objects were present

due to reﬂections being within the boundaries of

the track. The difference between distorted and

corrected point clouds decreased progressively

from the ﬁrst to the fourth quadrant (Renzler et al.,

2020).

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

314

• Vehicle Alignment. The trajectory planning re-

lies on a single mass model, which fails to con-

sider the vehicle alignment that may not always

be tangent to the path. This oversight results in

reduced acceleration during certain turns (Capo-

rale et al., 2019).

3.2 Indy Autonomous Challenge

The IAC conducted tests and races on various tracks,

including Indianapolis Motor Speedway, Las Vegas

Motor Speedway, Texas Motor Speedway, Lucas Oil

Raceway (Oval) and Monza Circuit. Initially, univer-

sity teams led the deployment of the AV-21, with the

ﬁrst shakedown occurring at Lucas Oil Raceway at

lower speeds, not reaching the vehicle’s dynamics’

peak performance because of the physical limitations

of the track itself.

Table 2 provides insight into vehicle speed perfor-

mance and race formats. Similar to Roborace, achiev-

ing higher speeds correlates directly with the avail-

ability of track time and space, allowing researchers

and engineers to simulate and enhance the software

stack’s robustness and validity. Numerous publica-

tions and issues have surfaced in connection with this

race series (Betz et al., 2022), highlighting challenges

in integrating the hardware stack, including extero-

ceptive failures and occasional powertrain issues. An-

ticipating such problems is crucial, as the technology

may encounter errors even under normal conditions,

and exposure to vibrations and lateral forces can fur-

ther accentuate sensor limitations.

Lidar faces challenges, including its high cost,

limitations in mechanical scanning, susceptibility to

disturbances from external light sources, and safety

constraints for the human eye, which curtail its de-

tection distance to approximately 100 meters (Wo-

jtanowski et al., 2014) .The LiDAR used for this

project was from Luminar, the Hydra model (Lumi-

nar, 2021). Various challenges and solutions associ-

ated with LiDAR functioning were identiﬁed:

• Delay due to Reﬂection. A high count of reﬂec-

tions induced a lag in the LiDAR perception pro-

cess, leading to complications in object recogni-

tion, particularly at high speeds (Betz et al., 2023)

• Banking Angle. The hardware underwent al-

terations to narrow its opening angle on straight

paths and broaden its ﬁeld of view (FOV) when

negotiating turns. This modiﬁcation was neces-

sary to address limitations in its vertical FOV, par-

ticularly in response to changing banking angles.

• Scanning Issue. A global positioning method re-

lying on LiDAR encounters difﬁculties in ﬁnding

a scan matching solution at high speeds. There-

fore, integration of two GPS measurements be-

came necessary. (Lee et al., 2023b)

• Driver Crash. The Lidar driver crashed during

the start-up process which caused the lidar to re-

port the last before (Frederick, 2023)

In the context of highly dynamic scenarios involv-

ing ground vehicles, there has been a paucity of aca-

demic research addressing the resilience of localiza-

tion systems under substantial lateral forces. Speciﬁ-

cally, within the domain of autonomous ground vehi-

cle racing, numerous teams relied on conventional lo-

calization methods, including Extended Kalman Fil-

ters (EKF) and Sensor Fusion, as well as pre-existing

packages like Autoware’s Robot Localization (Moore

and Stouch, 2014), which integrates wheel odometry

and Inertial Measurement Unit (IMU) data. Some at-

tempts have also been made to incorporate LiDAR

as a backup localization source. However, due to

computational demands, the reliability of LiDAR at

speeds exceeding 100 mph remains a concern. To

succinctly summarize, the The following sections de-

tail the challenges encountered and the successes

achieved on the racetrack:

• Data Filtering. Different edge cases can cause

inaccurate GPS data which caused early crashed

while testing, for instance one team crashed due

to when the GPS output data showed that the car

rotated 90 degrees between two data points (Fred-

erick, 2023).

• Vibrations. Multiple positioning degradation of

GNSS units due to strong vibration (Lee et al.,

2022) .

• Need of Cellular or Internet Connectivity. Lack

of cellular connectivity introduced several issues

such as the RTK would not receive the correction

values in some areas (Frederick, 2023), this in-

deed is a problem in remote testing track or spe-

ciﬁc areas of a large track

• ECU Latency. erroneous hard brake command

was initiated by a hardware Electronic Control

Unit (ECU) module, unrelated to the motion plan-

ner and controller (Raji et al., 2022).

• Tuning Issues. Speed was limited due to a ca-

ble that was attached to the powertrain the system

was not connected, limiting the speed. Addition-

ally, the controller requested full throttle during

race time, and there was oscillation of throttle due

to a non-ideal tuning of the turbocharges and mal-

function of its mechanic (Raji et al., 2023a)

A Qualitative Review of Full Sized Autonomous Racing Vehicle Sensors: A Case Study

315

Table 2: Real Time Speed Performance Benchmark.

Group Type Track Speed (kmh) Multiagent Series Car Type Source

TUM Course Formula E 150 No Roborace Devbot 1.0 (Stahl et al., 2019b)

Graz Mix Zala Zone 100 No Roborace Devbot 2.0 (Zubaca et al., 2020)

Graz Road Course Circuit de Croix 162 No Roborace Devbot 2.0 (Schratter et al., 2021)

Pisa Mix Zala Zone 60 No Roborace Devbot 2.0 (Massa et al., 2020)

TUM Road Course Modena 198 No Roborace Devbot 2.0 (Stahl and Diermeyer, 2021)

Roborace Long Strip Elvington Airﬁeld 280 No Roborace Robocar (Roborace, 2019)

TUM Oval IMS 241 No IAC AV-21 (Betz et al., 2023)

Euroracing Oval IMS 180 No IAC AV-21 (Raji et al., 2023a)

KAIST Oval IMS 147 No IAC AV-21 (Lee et al., 2023a)

KAIST Oval LOR 100 No IAC AV-21 (Lee et al., 2023a)

TUM Oval LVMS 270 Yes IAC AV-21 (Betz et al., 2023)

Euroracing Oval LVMS 272 No IAC AV-21 (Raji et al., 2023a)

Euroracing Oval LVMS 226 Yes IAC AV-21 (Raji et al., 2023a)

KAIST Oval TMS 205 Yes IAC AV-21 (Lee et al., 2023a)

KAIST Oval LVMS 248 No IAC AV-21 (Lee et al., 2023a)

KAIST Oval LVMS 212 Yes IAC AV-21 (Lee et al., 2023a)

MIT-PITT Oval IMS 217 No IAC AV-21 (Spisak et al., 2022)

Polimove Long Strip SSC 308 No IAC AV-21 (IAC, 2022)

KAIST Course Monza 200 No IAC AV-21 (Lee et al., 2023a)

Euroracing Course Monza 245 No IAC AV-21 (Raji et al., 2023b)

4 CONCLUSIONS

Both Roborace and IAC faced common challenges

in real-time performance, such as the need for pre-

cise sensor fusion, adaptation to high-speed dynam-

ics, and constant adjustments to hardware limitations.

The robustness of LiDAR and GNSS systems in ex-

treme conditions became a recurring theme. As the

autonomous racing landscape evolves, ongoing real-

time evaluation of these vehicles remains at the fore-

front, posing open questions for the assessment of

real-time frameworks.

This research highlights and identiﬁes recurring

patterns and challenges in past autonomous vehicle

racing platforms, and it is an extension of (Mar et al.,

2024). Speciﬁcally, it accentuates controllability in

two competitions involving full-sized racing vehicles

that achieved speeds surpassing 300 km/h. The anal-

ysis uncovers a trend suggesting an upward trajectory

in speed; however, there is no distinct surge in the

number of vehicles participating in multiagent rac-

ing scenarios. Across these studies, it is noteworthy

that no more than two vehicles have been subjected

to real high-speed racing competition. The qualitative

assessment of sensors is based on participant publi-

cations and publicly accessible data. A subsequent

study will explore diverse vehicle sizes, providing a

more quantitative analysis of sensor capabilities in

high-stress and off-road scenarios.

REFERENCES

A2RL (2023). Abu dhabi autonomous racing league. Ac-

cessed: Jan. 16, 2024.

Ahangar, M. N., Ahmed, Q. Z., Khan, F. A., and Hafeez, M.

(2021). A survey of autonomous vehicles: Enabling

communication technologies and challenges. Sensors,

21(3):706.

Ayala, R. and Mohd, T. K. (2021). Sensors in autonomous

vehicles: A survey. Journal of Autonomous Vehicles

and Systems, 1(3).

BBC (2016). Driverless roborace car speeds around don-

ington. https://www.bbc.com/news/av/37214619.

Accessed: Oct. 08, 2023.

Betz, J. et al. (2022). Autonomous vehicles on the

edge: A survey on autonomous vehicle racing. IEEE

Open Journal of Intelligent Transportation Systems,

3:458–488.

Betz, J. et al. (2023). Tum autonomous motorsport: An

autonomous racing software for the indy autonomous

challenge. J Field Robot, 40(4):783–809.

Betz, J., Wischnewski, A., Heilmeier, A., Nobis, F., Stahl,

T., Hermansdorfer, L., Lohmann, B., and Lienkamp,

M. (2019). What can we learn from autonomous level-

5 motorsport? In 9th International Munich Chassis

Symposium 2018: chassis. tech plus, pages 123–146.

Springer.

Buehler, M., Iagnemma, K., and Singh, S. (2009). The

DARPA urban challenge: autonomous vehicles in city

trafﬁc, volume 56. springer.

Caporale, D., Fagiolini, A., Pallottino, L., Settimi, A.,

Biondo, A., Amerotti, F., Massa, F., De Caro, S.,

Corti, A., and Venturini, L. (2018). A planning and

control system for self-driving racing vehicles. In

2018 IEEE 4th international forum on research and

technology for society and industry (RTSI), pages 1–

6. IEEE.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

316

Caporale, D., Settimi, A., Massa, F., Amerotti, F., Corti, A.,

Fagiolini, A., Guiggiani, M., Bicchi, A., and Pallot-

tino, L. (2019). Towards the design of robotic drivers

for full-scale self-driving racing cars. In 2019 In-

ternational Conference on Robotics and Automation

(ICRA), pages 5643–5649. IEEE.

Dow, J. (2017). Roborace pits (wo)man versus autonomous

racing machine around hong kong track. https://electr

ek.co/2017/12/13/roborace-woman-vs-autonomous-r

acing-machine/. Accessed: Oct. 05, 2023.

Fingas, J. (2018). Roborace’s self-driving car isn’t faster

than a human (yet). https://web.archive.org/web/2018

0621015907/https://www.engadget.com/2018/05/14/

roborace-self-driving-car-versus-human/. Accessed:

Oct. 05, 2023.

Frederick, R. (2023). Autonomous Vehicles Handling Anal-

ysis for an Indy Autonomous Challenge Car. PhD the-

sis, The University of Alabama, Tuscaloosa.

IAC (2020). Indy autonomous challenge. https://www.indy

autonomouschallenge.com/.

IAC (2022). Indy autonomous challenge racecar and ream

polimove set new land speed record for autonomous

racecar. [Accessed 22-02-2024].

Kelion, L. (2017). Driverless roborace car crashes at speed

in buenos aires. https://www.bbc.com/news/technol

ogy-39027477. Accessed: Oct. 08, 2023.

Knight, M. and Blendis, S. (2016). Driverless “roborace”

car makes street track debut in marrakech.

Lee, D., Jung, C., Finazzi, A., Seong, H., and Shim, D. H.

(2022). Resilient navigation and path planning system

for high-speed autonomous race car. arXiv preprint

arXiv:2207.12232.

Lee, D., Nam, H., Ryu, C., Nah, S., Moon, S., and Shim,

D. H. (2023a). Enhancing state estimator for au-

tonomous race car: Leveraging multi-modal system

and managing computing resources. arXiv preprint

arXiv:2308.07173.

Lee, D., Nam, H., Ryu, C., Nah, S., and Shim, D. H.

(2023b). Resilient navigation based on multimodal

measurements and degradation identiﬁcation for high-

speed autonomous race cars. In 2023 IEEE Intelligent

Vehicles Symposium (IV), page 1–8.

Luminar (2021). Hydra. https://levelfivesupplies.com/wp

-content/uploads/2020/08/Luminar-Hydra-Datasheet

.pdf. Accessed: Jan. 02, 2024.

Mar, M., Chellapandi, V., Yuan, L., Wang, Z., and Dietz,

E. (2024). A review of full-sized autonomous racing

vehicle sensor architecture.

Massa, F., Bonamini, L., Settimi, A., Pallottino, L., and

Caporale, D. (2020). Lidar-based gnss denied lo-

calization for autonomous racing cars. Sensors,

20(14):3992.

Moore, T. and Stouch, D. (2014). A generalized extended

kalman ﬁlter implementation for the robot operating

system. In Proceedings of the 13th International Con-

ference on Intelligent Autonomous Systems (IAS-13).

Springer.

NVIDIA (2016). Nvidia px2. https://developer.nvidia.com

/drive/px2.

On-Road Automated Driving (ORAD) Committee (2021).

Taxonomy and deﬁnitions for terms related to driving

automation systems for on-road motor vehicles. SAE

International.

Raji, A., Caporale, D., Gatti, F., Giove, A., Verucchi, M.,

Malatesta, D., Musiu, N., Toschi, A., Popitanu, S. R.,

Bagni, F., et al. (2023a). er. autopilot 1.0: The full

autonomous stack for oval racing at high speeds. arXiv

preprint arXiv:2310.18112.

Raji, A., Liniger, A., Giove, A., Toschi, A., Musiu, N.,

Morra, D., Verucchi, M., Caporale, D., and Bertogna,

M. (2022). Motion planning and control for multi ve-

hicle autonomous racing at high speeds. In 2022 IEEE

25th International Conference on Intelligent Trans-

portation Systems (ITSC), pages 2775–2782. IEEE.

Raji, A., Musiu, N., Toschi, A., Prignoli, F., Mascaro, E.,

Musso, P., Amerotti, F., Liniger, A., Sorrentino, S.,

and Bertogna, M. (2023b). A tricycle model to ac-

curately control an autonomous racecar with locked

differential. arXiv preprint arXiv:2312.14808.

Renzler, T., Stolz, M., Schratter, M., and Watzenig, D.

(2020). Increased accuracy for fast moving lidars:

Correction of distorted point clouds. In 2020 IEEE In-

ternational Instrumentation and Measurement Tech-

nology Conference (I2MTC), page 1–6.

Roborace (2016). Roborace. Accessed: Oct. 01, 2023.

Roborace (2019). The world’s fastest autonomous car robo-

car guinness world record. [Accessed 22-02-2024].

Sarkar, S. and Mohan, B. (2019). Review on Autonomous

Vehicle Challenges: AICC 2018, pages 593–603.

Schratter, M. et al. (2021). Lidar-based mapping and lo-

calization for autonomous racing. In Proc. Int. Conf.

Robot. Autom.(ICRA) Workshop Opportunities Chal-

lenges Auton. Racing, page 1–6.

Speedgoat (2016). Mobile real-time target machine. https:

//www.speedgoat.com/products-services/real-time-t

arget-machines/mobile-real-time-target-machine.

Spisak, J., Saba, A., Suvarna, N., Mao, B., Zhang, C. T.,

Chang, C., Scherer, S., and Ramanan, D. (2022). Ro-

bust modeling and controls for racing on the edge.

arXiv preprint arXiv:2205.10841.

Stahl, T. and Diermeyer, F. (2021). Online veriﬁcation en-

abling approval of driving functions—implementation

for a planner of an autonomous race vehicle. IEEE

Open Journal of Intelligent Transportation Systems,

2:97–110.

Stahl, T., Wischnewski, A., Betz, J., and Lienkamp, M.

(2019a). Multilayer graph-based trajectory planning

for race vehicles in dynamic scenarios. In 2019

IEEE Intelligent Transportation Systems Conference

(ITSC), page 3149–3154.

Stahl, T., Wischnewski, A., Betz, J., and Lienkamp, M.

(2019b). ROS-based localization of a race vehicle at

high-speed using lidar. In E3S Web of Conferences,

volume 95, page 04002. EDP Sciences.

Tumeo, A., Ceriani, M., Palermo, G., Minutoli, M., Castel-

lana, V., and Ferrandi, F. (2017). Chapter 3 - real-time

considerations for rugged embedded systems. In Vega,

A., Bose, P., and Buyuktosunoglu, A., editors, Rugged

A Qualitative Review of Full Sized Autonomous Racing Vehicle Sensors: A Case Study

317

Embedded Systems, pages 39–56. Morgan Kaufmann,

Boston.

Wojtanowski, J. et al. (2014). Comparison of 905 nm

and 1550 nm semiconductor laser rangeﬁnders’ per-

formance deterioration due to adverse environmental

conditions. Opto-Electronics Review, 22(3).

Yeong, D. J., Velasco-Hernandez, G., Barry, J., and Walsh,

J. (2021). Sensor and sensor fusion technology in au-

tonomous vehicles: A review. Sensors, 21(6):2140.

Zhu, Q. et al. (2021). Control informed design of the

IAC autonomous racecar for operation at the dynamic

envelope. In Proc. Int. Conf. Robot. Autom.(ICRA)

Workshop Opportunities Challenges Auton. Racing,

page 1–7.

Zubaca, J., Stolz, M., and Watzenig, D. (2020). Extended h

ﬁlter adaptation based on innovation sequence for ad-

vanced ego-vehicle motion estimation. In 2020 IEEE

3rd Connected and Automated Vehicles Symposium

(CAVS), pages 1–5. IEEE.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

318