Autonomous System: Human-Robot Interaction (HRI)

Transcription

1 Autonomous System: Human-Robot Interaction (HRI) MEEC MEAer 2014 / 2015! Course slides Rodrigo Ventura

2 Human-Robot Interaction (HRI) Systematic study of the interaction between humans and robots

3 Examples of sociable robots Kismet (MIT) Asimo (Honda) icub (RobotCub) hyper-realistic Einstein robot

4 Uncanny valley

5 Example: RAPOSA robot Tracked wheels robot designed for SAR operations SME+University joint project (IdMind+IST, ) thermal camera onboard sensors:! CO! methane! propane and butane! hydrogen sulphide! temperature! humidity! 2-axis accelerometer frontal cameras! +! illuminating lights b wireless link frontal arm length: 75cm, width: 37cm, height: 17.5cm and weight: 27Kg

6 Teleoperation user experience Situation awareness is one of the most important factors in SAR operations Conventional user interfaces show camera images in small scale Currently 2:1 human-to-robot ratio (Burke et al, 2004)

7 Example: RAPOSA interface console (laptop) wireless

8 Example: new RAPOSA interface (1) Head Mounted Display (HMD) provides 3D perception of the robot field of view

9 Example: new RAPOSA interface (2) Integrated tracker of the HMD allows movement of the robot accordingly to head orientation

10 Example: RAPOSA-NG interface Conventional GUI % batteries electronics motors 13.1 V 12.6 V 100 network name: raposa signal level: dbm bit rate: 54 Mb/s 0 % Wi-Fi connection N NW NE W heading body pitch body roll º º º E frontal angle SW SE S RAPOSA-NG HMD + Pan&Tilt º

11 Example: GermanOpen 2012

12 Example: CoBot (Carnegie Mellon Univ.)

13 Issues in HRI Human-Computer Interaction (HCI) design principles apply naturally to HRI Nielsen s heuristics applied to HRI - Is the robot s information presented in a way that makes sense to human controllers? - Can the human(s) control the robot(s) without having to remember information presented in various parts of the interface? - Is the interface consistent? - Does the interface provide feedback? - Does the interface have a clear and simple design?

14 Issues in HRI Nielsen s heuristics applied to HRI (cont.) - Does the interface help prevent, and recover, from errors? - Does the interface follow real-world conventions? - Is the interface forgiving; does it allow reversible actions? - Does the interface make it obvious what actions are available at any given point? - Does the interface provide shortcuts and accelerators?

15 Norman s HCI model seven stages of interaction: 1. formulation of the goal, in high-level terms 2. formulation of the intention, what will satisfy the goal? 3. specification of the action, what actions are necessary to accomplish goal? 4. execution of the action 5. perception of the system state, after/while executing action 6. interpretation of the system state 7. evaluation of the outcome, does it accomplish goal initially set? These stages ate iterated until goal achieved (or modified)

16 Evaluation methods Drawing from HCI methodologies: - inspection methods focused on experts e.g., compliance with Nielsen s heuristics - empirical methods focused on users e.g., usability testing, thinking aloud - formal methods focused on analytical methods e.g., time to perform certain primitive tasks

17 Example: HMD for RAPOSA Replacing traditional GUI with a Head-Mounted Display (HMD) in RAPOSA! 52! CHAPTER 5. RESULTS AND EVA yellow arena, orange arena and red arena, which provide di erent ch robots, such as random mazes, stairs, ramps, blocked means of acc Empirical evaluation using usability the questionaires Fig depicts the scenario built for Experiment III and highligh eral challenges which were incorporated into it, based on the NIST ref Formal evaluation measuring time to find 5.1. EXPERIMENT I DEPTH PERCEPTION 45objects arenas (the total area of the scenario is of approximately 50m2 ). - Figure 5.2: Scenario of Experiment I.

18 Usage evaluation Systematic evaluation of usage of the system! Ten subjects were used in the tests! Experiments conducted: - 1. depth perception to evaluate stereopsis - 2. detail perception effectiveness of large scale images - 3. timed execution of a task overall system evaluation

19 Experiment 1: depth perception

20 Experiment 2: detail perception

21 Experiment 3: task execution Inspired on the Reference Test Arena by NIST (Jacoff et al, 2003)

22 Experiment 3: task execution Task: from starting point, detect and take snapshots of objects of interest... M1 - using the GUI M2 - using the HMD M3 - M2 w/o yaw

23 A HRI taxonomy (1) Task type - e.g., urban search and rescue, walking aid for the blind, toy, etc.! (2) Task criticality - high, e.g., as in disaster situations - medium, e.g., hospital delivery robot - low, e.g., robot soccer

24 A HRI taxonomy (3) Robot morphology - anthropomorphic: human appearance, e.g., ASIMO humanoid - zoomorphic: non-human animal appearance, e.g., AIBO robot dog - functional: related to robot function, e.g., RAPOSA! (4) Ratio of people to robots - non-reduced fraction of number of humans over number of robots - e.g., in urban search and rescue 2/1 is a commonly found

25 A HRI taxonomy (5) Composition of robot teams - homogeneous, all of the same type, e.g., soccer team - heterogeneous, different types, e.g., land robot + aerial robot! (6) Level of shared information among teams - do humans coordinate themselves before issue commands, or do they operate independently? - do robots coordinate themselves indepently?

26 A HRI taxonomy - possible combinations of single and multiple humans and robots: A. H B. H C. H D. H H R R R R R R E. H H F. H H G. H H H. H H R R R R R R R

27 A HRI taxonomy (7) Interaction roles - supervisor interaction: setup and modification of goals and plans, e.g., referee box in robotic soccer - operator interaction: direct action-level interaction, e.g., tele-operated search and rescue robots - mechanic role: hardware modifications, e.g., Lego Mindstorms - peer interaction: coordination actions among humans and/or robots, cooperative behaviors among robotic soccer robots - bystander role: human (or robot) external to the robot loop of command, e.g., emotions and social interaction

28 A HRI taxonomy (8) Type of human-robot physical proximity - avoiding - passing - following - approaching - touching - none

29 A HRI taxonomy (9) Decision support for operators - available sensors: list of all available sensors, baseline for following fields: - provided sensors: list of sensor information provided to the operator - sensor fusion: list of sensor fusion mappings, e.g., {{sonar,laser} map} - pre-processing: list of sensor data processing operations! (10) Time/space taxonomy - in terms of time: synchronous (same time) or asynchronous (diff. times) - in terms of space: collocated (same place) or non-collocated (diff. places)

30 A HRI taxonomy (11) Autonomy level - autonomy: from 0% (teleoperated) to 100% (fully autonomous) - intervention: = 1 - autonomy - adjustable autonomy: the user is given the possibility of defining the robot s level of autonomy dynamically - example: the development of automated tasks for RAPOSA, namely automatic docking and automatic stair climbing, allow to increase the level of autonomy of the system

31 Situational awareness The knowledge of what is going on around the robot Critical in search and rescue operations time is crucial for victim survival

32 Situational awareness Levels of situation awareness: - level one: the perception of important information for the task at hand! - level two: the comprehension and/or integration of multiple pieces of information and to determine their relevance to the goals! - level three: the capability of forcasting future situation events and dynamics based on the perception of the present situation

33 References J. Scholtz, Theory and evaluation of human robot interaction, Proceedings of 36th Hawai International Conference on Systems Sciences (HICSS-03), H. A. Yanco, J. L. Drury, A taxonomy for human-robot interaction, Proceedings of the AAAI Fall Symposium on Human-Robot Interaction, H. A. Yanco, J. L. Drury, Classifying Human-Robot Interaction: An Updated Taxonomy, Proceedings of the IEEE Conference on Systems, Man and Cybernetics (SMC-04), A. Steinfeld, T. Fong, D. Kaber, M. Lewis, J. Scholtz, A. Scholtz, M. Goodrich, Common metric for human-robot interaction, Proceedings of Human Robot Interaction (HRI-06), H. Martins, R. Ventura, Immersive 3-D Teleoperation of a Search and Rescue Robot Using a Head-Mounted Display, Proceedings of 14th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA-2009), 2009.!