PUMA: Fully Decentralized Uncertainty-aware Multiagent Trajectory Planner with Real-time Image Segmentation-based Frame Alignment

TL;DR

An uncertainty-aware multiagent trajectory planner and an image segmentation-based frame alignment pipeline are presented that rectifies inter-agent frame misalignment and ensures safe navigation in unknown environments and collision avoidance in decentralized settings.

Abstract

Fully decentralized, multiagent trajectory planners enable complex tasks like search and rescue or package delivery by ensuring safe navigation in unknown environments. However, deconflicting trajectories with other agents and ensuring collision-free paths in a fully decentralized setting is complicated by dynamic elements and localization uncertainty. To this end, this paper presents (1) an uncertainty-aware multiagent trajectory planner and (2) an image segmentation-based frame alignment pipeline. The uncertainty-aware planner propagates uncertainty associated with the future motion of detected obstacles, and by incorporating this propagated uncertainty into optimization constraints, the planner effectively navigates around obstacles. Unlike conventional methods that emphasize explicit obstacle tracking, our approach integrates implicit tracking. Sharing trajectories between agents can cause potential collisions due to frame misalignment. Addressing this, we introduce a novel frame alignment pipeline that rectifies inter-agent frame misalignment. This method leverages a zero-shot image segmentation model for detecting objects in the environment and a data association framework based on geometric consistency for map alignment. Our approach accurately aligns frames with only 0.18 m and 2.7 deg of mean frame alignment error in our most challenging simulation scenario. In addition, we conducted hardware experiments and successfully achieved 0.29 m and 2.59 deg of frame alignment error. Together with the alignment framework, our planner ensures safe navigation in unknown environments and collision avoidance in decentralized settings.

Figures (7)

• Figure 1: PUMA is thoroughly evaluated in simulation and hardware experiments are also performed to evaluate the real-time image segmentation-based frame alignment pipeline.
• Figure 2: Frame Alignment Pipeline Workflow: The process starts by un-distorting a raw, fisheye image. Next, the pipeline identifies and segments individual objects within the image. We then determine each identified object's centroid, filtering out objects based on size. Once processed, these centroids are projected onto a 2D map, which is used in the frame alignment among agents.
• Figure 3: PUMA balances reducing uncertainties of (1) known obstacles and (2) potential obstacles in trajectory: The green pyramid represents the *fov. The agent navigates around the dynamic obstacle to reach its destination on the opposite side. The trajectory color shows the agent's velocity — red is fast, and blue slow.
• Figure 4: Pipeline Evaluation Simulation Environments.
• Figure 5: Frame alignment quality in Case 3 (difficulty: Easy): Trajectory color shows the estimation error associated with it and the color of the coordinate frame attached to vehicle 1 shows the yaw estimation error. The pipeline successfully estimates the drifted state of vehicle 1 with minimal errors (Table \ref{['tab:sim-benchmarking-pads']}). The coordinate frames are captured at intervals of 5.