Not an Obstacle for Dog, but a Hazard for Human: A Co-Ego Navigation System for Guide Dog Robots
Authors: Ruiping Liu, Jingqi Zhang, Junwei Zheng, Yufan Chen, Peter Seungjune Lee, Di Wen, Kunyu Peng, Jiaming Zhang, +3 more
Organizations: The authors are with Karlsruhe Institute of Technology, 76131 Karlsruhe, Germany. · The author is with ETH Zurich, 8092 Zurich, Switzerland. · The author is with INSAIT, Sofia University “St. Kliment Ohridski”, 1784 Sofia, Bulgaria. · The authors are with Hunan University, Changsha 410012, China.
Guide dogs offer independence to Blind and Low-Vision (BLV) individuals, yet their limited availability leaves the vast majority of BLV users without access. Quadruped robotic guide dogs present a promising alternative, but existing systems rely solely on the robot's ground-level sensors for navigation, overlooking a critical class of hazards: obstacles that are transparent to the robot yet dangerous at human body height, such as bent branches. We term this the viewpoint asymmetry problem and present the first system to explicitly address it. Our Co-Ego system adopts a dual-branch obstacle avoidance framework that integrates the robot-centric ground sensing with the user's elevated egocentric perspective to ensure comprehensive navigation safety. Deployed on a quadruped robot, the system is evaluated in a controlled user study with sighted participants under blindfold across three conditions: unassisted, single-view, and cross-view fusion. Results demonstrate that cross-view fusion significantly reduces collision times and cognitive load, verifying the necessity of viewpoint complementarity for safe robotic guide dog navigation.
Figures & tables
Fig. 1 : Corner cases where the human encounters hazards that are not obstacles for the guide dog [ 6 ] .
Fig. 2 : Hardware components of our Co-Ego system.
Fig. 3 : Viewpoint asymmetry between human and robot egocentric views. (a) RGB and (b) point cloud from the user’s neck-worn camera capture the hanging lamp at head height but miss the ground-level cabinet. (c) RGB and (d) cost map from the quadruped’s point cloud detect ground-level obstacles (red regions) with a navigable path in between (blue regions) but remain blind to the overhead lamp.
Fig. 4 : Change of velocity and turns when avoiding an obstacle. vx denotes the forward velocity, while wz denotes the angular velocity. Zone I corresponds to turning to avoid the obstacle, whereas Zone II corresponds to recovery.
Fig. 5 : Flowchart of the dual-perception priority-based navigation algorithm.
Fig. 6 : Flowchart of the environment warning process.
Fig. 7 : Test scenario. The hallway includes three ground-level obstacles affecting both the robot dog and the human (a chair and two cabinets), and two hanging obstacles affecting only the human (a lamp and a bent branch).
Fig. 8 : Illustration of the three navigation conditions evaluated in our study.
Fig. 9 : Quantitative comparison of the three navigation conditions: without assistance, with robot guide dog, and with our Co-Ego system. From left to right, the plots show (a) number of collisions, (b) task completion time, and (c) perceived safety. Error bars denote participant variability.
Fig. 10 : NASA-TLX cognitive load measured while completing tasks under three conditions: without assistance , with the robot guide dog , and with our Co-Ego system . Lower scores indicate lower cognitive load.
Electronics and Telecommunications Research Institute (ETRI), Korea · School of Computer Science and Engineering, Pusan National University, Korea · Division of Electronics and Information Engineering, Jeonbuk National University, Korea