Decentralized Multi-Agent Trajectory Planning in Dynamic Environments with Spatiotemporal Occupancy Grid Maps

TL;DR

A decentralized trajectory planning framework for the collision avoidance problem of multiple micro aerial vehicles (MAVs) in environments with static and dynamic obstacles, which extends the kinodynamic A* and the corridor-constrained trajectory optimization algorithms to efficiently tackle static and dynamic obstacles with arbitrary shapes.

Abstract

This paper proposes a decentralized trajectory planning framework for the collision avoidance problem of multiple micro aerial vehicles (MAVs) in environments with static and dynamic obstacles. The framework utilizes spatiotemporal occupancy grid maps (SOGM), which forecast the occupancy status of neighboring space in the near future, as the environment representation. Based on this representation, we extend the kinodynamic A* and the corridor-constrained trajectory optimization algorithms to efficiently tackle static and dynamic obstacles with arbitrary shapes. Collision avoidance between communicating robots is integrated by sharing planned trajectories and projecting them onto the SOGM. The simulation results show that our method achieves competitive performance against state-of-the-art methods in dynamic environments with different numbers and shapes of obstacles. Finally, the proposed method is validated in real experiments.

Figures (6)

• Figure 1: 4 robots navigate in a dynamic environment with 20 columns and 20 circles. All robots perform independent planning and coordinate with trajectory sharing. The spatio-temporal corridors are indicated in blue, and obstacle velocities are marked by red arrows.
• Figure 2: Overview of the proposed framework for decentralized multi-robot trajectory planning.
• Figure 3: Performance comparison between our proposed method ( Ours), and two baselines ( MADER and EGO-Swarm) in 3 different tasks (Fig. \ref{['fig:case1']}--\ref{['fig:case3']}) and 2 different obstacle settings: mixed environments with both columns and circles (Fig. \ref{['fig:env1:case1:rst']}--\ref{['fig:env1:case3:rst']}), and pure column environments (Fig. \ref{['fig:env2:case1:rst']}--\ref{['fig:env2:case3:rst']}). In Fig. \ref{['fig:case1']}--\ref{['fig:case3']}, trajectories for each MAV are displayed with distinct colors. Fig. \ref{['fig:env1:case1:rst']}--\ref{['fig:env2:case3:rst']} evaluate the average flight time to the goal and failure rate with different numbers of obstacles, distinguishing failure types between deadlock and collisions.
• Figure 4: The computation time of each step in our method under different obstacle settings.
• Figure 5: (a) Composite image of two MAVs flying in an indoor dynamic environment. (b) The SOGM at $t=2.5s$ is displayed, showing the predicted future occupancies with a gradient transition from green to blue.