actual five 8x8 km
2
regions in Peru (.a.Delta,
b.Colorado, c.Madre de Dios, d.Inambari, e. La
Pampa) (Fig.4). We simulate motion of the UAV (as
well as the onboard object detection system) and keep
its altitude fixed by setting the field of view to 200x200
m
2
. We test our RDE approach against the AEP
approach developed in (Selin et al., 2019). However,
due to the space limitation, we only showed the heat
maps for two regions b and e (Colorado, La Pampa).
Fig.5 (a) shows the likelihood of ASGM areas in
the regions b and e shown in (Fig. 4), the color scale
is between yellow and green such that dark yellow
areas have higher probability of having ASGM. Fig.
5 (b) shows the most frequent explored positions in
region b using the proposed RDE approach. We
collected those points by running RDE on 100 trials
with 2000 time steps per each trail starting from
random positions in each run. The green and yellow
colors represent the most visited areas such that areas
in yellow are visited more than areas with green color.
We then explored the same region b using the AEP
(Fig. 4 (c)). The testing results for region e are
illustrated in Fig. 5. For both regions, our approach
was clearly able to navigate the majority of ASGM
areas in comparison to AEP while spending less time
inside vegetation areas. However, AEP was faster in
making decisions than RDE by average of 29% when
exploring the areas shown in Fig.4. AEP uses a
greedy algorithm which guaranteed faster execution
but not necessarily good coverage for ASGM while
RDE needs to compute the robustness of P-MTL
constraints before each exploration decision and use
the MCMC sampler to select the next target with
higher robustness.
Fig.7 illustrates the average coverage of ASGM in
all regions shown in Fig.4 using our RDE and AEP
with different numbers of time steps respectively. The
time steps here represent the battery life for the UAV.
The percentage of coverage grows linearly with the
allotted time for both approaches, but the RDE covers
more ASGM areas by approximately 38% over AEP.
6 CONCLUSIONS
In this paper, we presented a new exploration
approach RDE that incorporates the online discovered
knowledge into the exploration decisions for UAVs.
RDE uses the robustness of P-MTL specifications to
guide the stochastic process of MCMC to make the
exploration decisions in completely unknown
environment. We have tested our approach on four
simulated areas in Amazon forest in Peru to look for
mining areas (e.g. ASGM). In comparison to a greedy
approach called AEP (Selin et al., 2019), our
approach leads the UAV into more areas classified as
ASGM than AEP without getting stuck or spending
long time in vegetation areas. In future work, we
intend to test our approach on real UAVs in Amazon
forest. In order to do that, we have to incorporate the
dynamics of the UAV and the control information
(i.e. speed, altitude) into the P-MTL specifications of
the problem.
REFERENCES
A. Bircher, M. Kamel, K. Alexis, H. Oleynikova, and R.
Siegwart (2016). Receding horizon "Next-Best-View"
planner for 3D exploration. In Proceedings of
International Conference on Robotics and Automation
ICRA, 2016.
A. Dokhanchi, B. Hoxha, and G. Fainekos (2014), On-Line
monitoring for temporal logic robustness.In
Proceedings of International Conference on Runtime
Verification, 2014.
A.I.M Ayala, S.B. Andersson, and C. Belta (2013).
Temporal logic motion planning in unknown
environments. In IEEE/RSJ International Conference
on Intelligent Robots and Systems, 2013.
B. Adler, J. Xiao, and J. Zhang (2014), Autonomous
exploration of urban environments using unmanned
aerial vehicles. Journal of Field Robotics, 2014.
B. Yamauchi (1997). A frontier-based approach for
autonomous exploration. In Proceedings of IEEE
International Symposium on Computational
Intelligence in Robotics and Automation CIRA, 1997.
C. Vasile, J. Tumova, S. Karaman, C. Belta, and D. Rus
(2017). Minimum-violation scLTL motion planning for
mobility-on-demand. In Proceedings of the IEEE
International Conference on Robotics and Automation
(ICRA), 2017
F. S. Barbosa, D. Duberg, P. Jensfelt, and J. Tumova
(2019). Guiding autonomous exploration with signal
temporal logic. In Proceedings of IEEE Robotics and
Automation, 2019.
G. Asner, W. Llactayo, R. Tupayachi, E. Ráez-Luna (2013).
Elevated rates of gold mining in the Amazon revealed
through high-resolution monitoring. In Proceedings of
the National Academy of Sciences of the United States
of America, 2013.
G. Fainekos, and G. Pappas (2006), Robustness of temporal
logic specifications. In Proceedings of the First
combined international conference on Formal
Approaches to Software Testing and Runtime
Verification. 2006.
H. Abbas, G. Fainekos, S. Sankaranarayanan, F. Ivan, and
A. Gupta (2013). Probabilistic Temporal Logic
Falsification of Cyber-Physical Systems. ACM
Transactions on Embedded Computing Systems
(TECS), 2013.