Robotic Guide Dog: Leading a Human with Leash-Guided Hybrid Physical Interaction

Anxing Xiao, Wenzhe Tong, Lizhi Yang, Jun Zeng, Zhongyu Li, Koushil Sreenath

Seminal March 2021

TL;DR A Mini Cheetah quadruped guides a blindfolded person on a leash, and its planner lets the leash go slack so the pair can fit through spaces a taut 1.6 m robot-human system cannot. Three adults were guided through doorways and a corridor as narrow as 1.0 m without any collision.

What they do

Most robotic guides for visually impaired people are wheeled platforms with large bases or rigid actuated canes, which cannot fit through narrow spaces. This paper uses a small quadruped (a Mini Cheetah) with a leash instead. A slack leash lets the robot change the size of the robot-human system, so the planner can deliberately switch between taut and slack to get through doorways.

How it works

Hybrid model. When the leash is taut, the human is assumed to move in the direction of the leash force; when it is slack, the human stops and only the robot moves. Four discount coefficients, identified offline from guiding runs as [0.8, 0.8, 0.6, 0.8], absorb the human pulling back on the robot.
Leash tension model. A linear fit maps the robot’s speed along the leash to tension. The planner uses it as a two-sided constraint (fit plus or minus one standard deviation) instead of an equality.
Planning. An A* global planner gives waypoints; a mixed-integer collocation local planner (CasADi, IPOPT) chooses the taut/slack mode per step, with obstacle-avoidance constraints for both robot and human.
Perception and control. LiDAR SLAM and AMCL localize the robot; a depth camera on a 2-DoF gimbal tracks the human with a Kalman filter. MPC plus WBIC track the velocity commands.

Results

The taut-mode model predicts robot position with 0.023 m average error and human position with 0.176 m.
76% of measured tension samples fall within one standard deviation of the linear fit (σ = 15.06 N).
Three adults were guided to several goals in a map whose narrowest region is 1.0 m wide; all runs reached the goal without collision.
In one 75 s run, the leash went slack at the second doorway from 45 s to 60 s, tension dropped near zero and the human slowed below 0.05 m/s, until the robot repositioned and pulled taut again.

Limitations

The human model is simple (move along the taut leash, stop when slack); the authors list richer modeling of human behavior as future work. Tests used blindfolded sighted adults rather than visually impaired users, and the paper does not report how many guiding runs were made.

Videos

Figures

From Figure 1: The Mini Cheetah guides a blindfolded person with a leash that can be taut or slack. A 2D LiDAR locali... — From Figure 1: The Mini Cheetah guides a blindfolded person with a leash that can be taut or slack. A 2D LiDAR localizes the robot and a depth camera on a gimbal detects the human.

From Figure 3: The two hybrid modes of the robot-leash-human model. The system switches from slack to taut when the l... — From Figure 3: The two hybrid modes of the robot-leash-human model. The system switches from slack to taut when the leash tightens, and back to slack when it loosens.

From Figure 4: The guide-dog framework. An A* global planner gives waypoints to the local planner with physical human... — From Figure 4: The guide-dog framework. An A* global planner gives waypoints to the local planner with physical human-robot interaction, which sends velocity commands to the Mini Cheetah tracking controller.

From Figure 7: Start of a guiding run from an initial to a target position. On the map, blue marks the robot, orange ... — From Figure 7: Start of a guiding run from an initial to a target position. On the map, blue marks the robot, orange the human and their past trajectory, and green the global planned reference.

Abstract

An autonomous robot that is able to physically guide humans through narrow and cluttered spaces could be a big boon to the visually-impaired. Most prior robotic guiding systems are based on wheeled platforms with large bases with actuated rigid guiding canes. The large bases and the actuated arms limit these prior approaches from operating in narrow and cluttered environments. We propose a method that introduces a quadrupedal robot with a leash to enable the robot-guiding human system to change its intrinsic dimension (by letting the leash go slack) in order to fit into narrow spaces. We propose a hybrid physical Human-Robot Interaction model that involves leash tension to describe the dynamical relationship in the robot-guiding human system. This hybrid model is utilized in a mixed-integer programming problem to develop a reactive planner that is able to utilize slack-taut switching to guide a blind-folded person to safely travel in a confined space. The proposed leash-guided robot framework is deployed on a Mini Cheetah quadrupedal robot and validated in experiments.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.

Matt Strong