MakeWay: Object-Aware Costmaps for Proactive Indoor Navigation Using LiDAR

TL;DR

The paper addresses proactive indoor robot navigation under dynamic clutter by introducing a LiDAR-based object-level SLAM that builds object-centric maps and affordance-aware costmaps to guide planning. It integrates detection, pose refinement, and tracking of objects with data association and CAD-model alignment into a ROS-based navigation stack, enabling proactive collision avoidance. A automatic indoor LiDAR labeling pipeline using CAD models is also proposed to address data scarcity. Across simulations and real-world experiments, the method improves safety (reduced risk time) while maintaining competitive efficiency and runs in real time onboard, with public code and data planned for release.

Abstract

In this paper, we introduce a LiDAR-based robot navigation system, based on novel object-aware affordance-based costmaps. Utilizing a 3D object detection network, our system identifies objects of interest in LiDAR keyframes, refines their 3D poses with the Iterative Closest Point (ICP) algorithm, and tracks them via Kalman filters and the Hungarian algorithm for data association. It then updates existing object poses with new associated detections and creates new object maps for unmatched detections. Using the maintained object-level mapping system, our system creates affordance-driven object costmaps for proactive collision avoidance in path planning. Additionally, we address the scarcity of indoor semantic LiDAR data by introducing an automated labeling technique. This method utilizes a CAD model database for accurate ground-truth annotations, encompassing bounding boxes, positions, orientations, and point-wise semantics of each object in LiDAR sequences. Our extensive evaluations, conducted in both simulated and real-world robot platforms, highlights the effectiveness of proactive object avoidance by using object affordance costmaps, enhancing robotic navigation safety and efficiency. The system can operate in real-time onboard and we intend to release our code and data for public use.

Figures (11)

• Figure 1: Incorporating semantic information for proactive collision avoidance, our system runs our object-level SLAM system and generates object-aware costmaps for indoor navigation.
• Figure 2: System architecture diagram. LiDAR input is consumed to produce 3D bounding boxes for chairs and tables. Their poses are then refined with ICP using their corresponding CAD models, and tracked across multiple frames. Next, our object-aware costmaps are attached to the tracked objects, which the planner uses to achieve object-aware proactive navigation.
• Figure 3: Object-aware costmaps. Our object-level dense mapping is on the left, and the right is the obtained costmap. Based on the affordance of chairs, we assign our elliptical cost aligned with the chairs' detected orientation.
• Figure 4: Our six-step data annotation pipeline. LiDAR point clouds are collected by the robot. The points are then classified, keeping the Movable points corresponding to occasionally moved objects (i.e., tables and chairs). Objects are then isolated using clustering, and rough 3D boxes are drawn around them, providing initializations for ICP-based CAD alignment. Finally, all Movable points are semantically classified as tables or chairs.
• Figure 5: Chair (left) and table (right) costmaps for Experimental Setup.