Gap Based Elastic Trees as a Novel Approach for Fast and Reliable Obstacle Avoidance for UGVs

This paper presents a novel approach to improve the online implementation performance of existing obstacle avoidance and path planning methods for environments with no maps. The Gap Based Elastic Trees (GET) algorithm proposed in this paper, combines some features of popular algorithms in path planning, obstacle avoidance, and path smoothing, and outperforms popular obstacle avoidance approaches designed for cluttered environments, also paving the way for the use of path planners in real-time applications. The design steps of GET consists of the following steps: gap calculation, smooth path generation and trajectory planning. Even though the GET method is an obstacle avoidance method, to demonstrate its efficiency we compared with offline version of base planner (RRT) of GET and a well-known shortest path algorithm (A*). The results demonstrate that, in most cases, the GET method guides the robot on almost the shortest routes with no requirement of an environment map. The GET algorithm also outperforms the obstacle avoidance methods with proven success in cluttered environments in terms of speed and safety. Further analysis was done to see the performance of the GET method in real world scenarios. These tests both validated simulation results and revealed that GET algorithm has more smooth trajectories than the most successful alternative. Overall, the improved performance of GET can be attributed to its predictive feature, and capability to plan ahead. Additionally, GET method can be adapted both robotic and autonomous car navigation tasks.

Automated summary

This paper presents a novel approach, called Gap Based Elastic Trees (GET), which improves the online implementation performance of obstacle avoidance and path planning methods for map-less environments. The GET algorithm combines features of popular algorithms in path planning, obstacle avoidance, and path smoothing, outperforming existing obstacle avoidance approaches and enabling real-time applications of path planners. The improved performance of GET is attributed to its predictive feature and capability to plan ahead. Additionally, the GET method can be adapted for both robotic and autonomous car navigation tasks.