Abstract:
Safe and efficient navigation is among the key challenges for sidewalk delivery
robots operating in urban environments. This paper proposes a hybrid pathplanning approach using the global planning capability of the A* algorithm
combined with local obstacle avoidance using the Artificial Potential Field
(APF) method.
Performance tests were carried out in simulated urban scenarios with a
wide range of algorithms. The two most promising ones were A* and APF.
A* brought computation time down to 1.2 seconds, only 8% of the total simulation time for the computation of a 12.4 km path. It outperforms Dijkstra,
Greedy, BFO, PSO, ACO, and GWO in both path efficiency and cost. APF
managed to avoid obstacles in 5.2 seconds for a 30-minute simulation video,
while other approaches required between 40 and 100 seconds.