Joint Localization and Planning Using Diffusion

Diffusion models have been successfully applied to robotics problems such as manipulation and vehicle path planning. In this work, we explore their application to end-to-end navigation - including both perception and planning - by considering the problem of jointly performing global localization and path planning in known but arbitrary 2D environments. In particular, we introduce a diffusion model which produces collision-free paths in a global reference frame given an egocentric LIDAR scan, an arbitrary map, and a desired goal position. To this end, we implement diffusion in the space of paths in $\text{SE}(2)$, and describe how to condition the denoising process on both obstacles and sensor observations. In our evaluation, we show that the proposed conditioning techniques enable generalization to realistic maps of considerably different appearance than the training environment, demonstrate our model's ability to accurately describe ambiguous solutions, and run extensive simulation experiments showcasing our model's use as a real-time, end-to-end localization and planning stack.

Authors

• Lukas Lao Beyer
• Sertac Karaman

Keywords

• Location awareness
• Training
• Laser radar
• Navigation
• Noise reduction
• Diffusion models
• Robot sensing systems
• Path planning
• Real-time systems
• Planning
• Denoising
• Diffusion Model
• Global Plan
• Global Frame
• Path In Space
• Global Localization
• Goal Position
• 2D Environment
• Problem In Robotics
• Sensor Observations
• Global Reference Frame
• Arbitrary Mapping
• Feature Maps
• Diffusion Process
• Control Input
• Brownian Motion
• Shortest Path
• Kernel Density
• Forward Process
• Environment Map
• Obstacle Avoidance
• Synthetic Environment
• Synthetic Examples
• Rotated Component
• Previous Solution