HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation

Human-aware space sharing and navigation for an interactive robot

Harmish Khambhaita 1
1 LAAS-RIS - Équipe Robotique et InteractionS
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : The methods of robotic movement planning have grown at an accelerated pace in recent years. The emphasis has mainly been on making robots more efficient, safer and react faster to unpredictable situations. As a result we are witnessing more and more service robots introduced in our everyday lives, especially in public places such as museums, shopping malls and airports. While a mobile service robot moves in a human environment, it leaves an innate effect on people about its demeanor. We do not see them as mere machines but as social agents and expect them to behave humanly by following societal norms and rules. This has created new challenges and opened new research avenues for designing robot control algorithms that deliver human-acceptable, legible and proactive robot behaviors. This thesis proposes a optimization-based cooperative method for trajectoryplanning and navigation with in-built social constraints for keeping robot motions safe, human-aware and predictable. The robot trajectory is dynamically and continuously adjusted to satisfy these social constraints. To do so, we treat the robot trajectory as an elastic band (a mathematical construct representing the robot path as a series of poses and time-difference between those poses) which can be deformed (both in space and time) by the optimization process to respect given constraints. Moreover, we also predict plausible human trajectories in the same operating area by treating human paths also as elastic bands. This scheme allows us to optimize the robot trajectories not only for the current moment but for the entire interaction that happens when humans and robot cross each other's paths. We carried out a set of experiments with canonical human-robot interactive situations that happen in our everyday lives such as crossing a hallway, passing through a door and intersecting paths on wide open spaces. The proposed cooperative planning method compares favorably against other stat-of-the-art human-aware navigation planning schemes. We have augmented robot navigation behavior with synchronized and responsive movements of the robot head, making the robot look where it is going and occasionally diverting its gaze towards nearby people to acknowledge that robot will avoid any possible collision with them. At any given moment the robot weighs multiple criteria according to the social context and decides where it should turn its gaze. Through an online user study we have shown that such gazing mechanism effectively complements the navigation behavior and it improves legibility of the robot actions. Finally, we have integrated our navigation scheme with a broader supervision system which can jointly generate normative robot behaviors such as approaching a person and adapting the robot speed according to a group of people who the robot guides in airports or museums.
Document type :
Complete list of metadata

Contributor : Abes Star :  Contact
Submitted on : Tuesday, May 4, 2021 - 10:17:48 AM
Last modification on : Monday, April 4, 2022 - 3:24:37 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01931244, version 2


Harmish Khambhaita. Human-aware space sharing and navigation for an interactive robot. Artificial Intelligence [cs.AI]. Université Paul Sabatier - Toulouse III, 2017. English. ⟨NNT : 2017TOU30399⟩. ⟨tel-01931244v2⟩



Record views


Files downloads