The safe & productive robot working without fences

Transcription

1 The European Robot Initiative for Strengthening the Competitiveness of SMEs in Manufacturing The safe & productive robot working without fences Final Presentation, Stuttgart, May 5 th, 2009

2 Objectives (1) Assessment and review of the state of the art of safety sensors for human detection, motion tracking, etc. (2) Functional and technical specification of safety compliant sensors and sensing principles for human tracking and motion detection (3) Development of safety compliant sensing for human tracking and motion detection (4) Design and development of sensor-based reaction and navigation strategies (5) Improvements, integration of sensing and robot control into set up (extracted from the Description of Work, the contract between the consortium and the EC) 2

3 1. Concept Introduction Outline 2. Safety sensors for the human detection and tracking 3. Reaction strategies to avoid collision and increase productiveness 4. Application scenarios 3

4 The safe & productive robot working without fences 1. Concept Introduction 4

5 Safety Analysis and Certification HUMAN ROBOT INTERACTION Cooperation Work Space Sharing ISO

6 Safety Analysis and Certification HUMAN ROBOT INTERACTION Cooperation Work Space Sharing ISO on going ISO :2009 ISO :2006 Robots for industrial environments Safety requirements -1 Robots a)safe reduced speed and enabling device. b) The robot shall maintain a separation distance from the operator with safe reduced speed and safe monitored position c) The robot s power and force is limited by inherent design or a safe control system [ 80 W / 150 N ]. 6

7 Safety Analysis and Certification HUMAN ROBOT INTERACTION Cooperation Work Space Sharing ISO on going ISO :2009 ISO :2006 Robots for industrial environments * Risk Analysis Safety requirements Risk -1 Robots * Risk Evaluation Assessment Risk * Risk Control Management * Information from phases after production a)safe reduced speed and enabling device. b) The robot shall maintain a separation distance from the operator with safe reduced speed and safe monitored position c) The robot s power and force is limited by inherent design or a safe control system [ 80 W / 150 N ]. IEC and IEC

8 Applications & SMErobot requirements 1. Which applications allow a robot working without fences? 2. Which applications allow the trajectory re-planning or execution velocity modification? 8

9 The safe & productive robot working without fences 2. Safety sensors for the human detection and tracking 9

10 SENSORS & DETECTION Where should detection sensors be installed? FIXED TO THE ROBOT (Local Approach) FIXED TO THE WORLD (Global Approach) Merging of different sensors, fixed to the robot and to the world 10

11 SENSORS & DETECTION: Global Approach Robot Cell Supervision (1)TECHNOLOGICAL PROBLEMS Resolution, Safe sensors, Real-time communication, Safe communication, Illuminations, (2)APPLICATION DEPENDANT PROBLEMS Overlapping, Tracking of human motion, Minimum distance allowed, NEW SENSORS OFF THE SHELVES SENSORS Time of Flight Camera Cheap 2D Camera 11

12 Global Approach: NEW SENSORS Time Of Flight Camera - 3D IMAGE 3D Sensor data processing: -> Identification of robot -> Identification of obstacles -> Intrusion detection for safe areas around robot links Camera on the top of the workcell Used in the woodworking workcell (hard and dusty environment) Used in the bin-picking workcell (AUTOMATICA fair) under certification process by BGIA - SISL III 12

13 Global Approach: OFF THE SHELVES SENSORS CHEAP 2D CAMERA and ADVANCED ALGORITHMS -> Markov model for the TRACKING of MULTIPLE HUMANS in user defined HOTZONES -> environment/background subtraction -> markers management -> not yet real-time PRODUCTS ARE ON THE MARKET. TrueView People Counter. TrueView Hotzone for adjacent application (close environment supervision) 13

14 SENSORS & DETECTION: Local Approach SENSORS FIXED TO THE ROBOT (1) TECHNOLOGICAL PROBLEMS Resolution, Safety, Real-time communication, Safe communication, Illuminations (2) APPLICATION DEPENDANT PROBLEMS Correlation ACQUISITION CONFIGURATION, Difficult Tracking of human motion, Minimum distance allowed NEW SENSORS OFF THE SHELVES SENSORS Capacitive Sensors Laser scanner, Force Torque Sensor 14

15 SENSORS & DETECTION: Local Approach Measurement DEPENDS ON the robot/application configuration Sensor Robot Measured DATA Status/Stop COMPARE UNIT ESTIMATED VALUE IN ABSENCE OF OBSTACLE Sensor database Different sensors: - For the links - For the tool Still a prototypes Good performances HOWEVER technological problems still active Internal test-bed (available only for internal operators) 15

16 (50x64 pixel, 60 ms) SENSORS & DETECTION MERGE OF DIFFERENT SOURCES Algorithms for tracking multiple humans Local Approach Time of Flight Camera Cheap 2D Camera Capacitive Sensors Algorithms to merge different sensors signals and obstacles tracking BAYESIAN OCCUPANCY GRID 16

17 The safe & productive robot working without fences 3. Reaction strategies to avoid collision and increase productiveness 17

18 REACTION STRATEGIES How react when the obstacles is detected? ONLY VELOCITY MODIFICATION TRAJECTORY ON- LINE REPLANNING Merge of different strategies 18

19 REACTION STRATEGIES HIGH LEVEL Logic control for collision avoidance How define the safe area, warning area,? ** APPLICATION DEPENDANT ** KINETIC ENERGY of the robot is the main source of potential dangerous ** CLAMPING of humans arms or hands ** INSTRUCTION LEVEL of the OPERATORS 19

20 REACTION STRATEGIES: Velocity control INTRUSION DETECTION COLLISION WITHDRAW Normal direction of movement IF intrusion THEN stop IF collision foreseen THEN. Slow down. Stop. Apply a stored reverse movement Evasive movement CONTINOUS AND FUZZY REGULATION IF close AND NOT far THEN slow down quickly 20

21 REACTION STRATEGIES: Dynamic Behavior Control INDUSTRIAL ARTIFICIAL POTENTIAL FIELD Description of environment: Obstacles Target FIELD of Attractive Forces FIELD of Repulsive Forces The Resultant Force Field is applied to the robot kinematic structure and the path is automatically defined The trajectory follow the minimum potential direction Global Optimum 21

22 REACTION STRATEGIES: Dynamic Behavior Control INDUSTRIAL ARTIFICIAL POTENTIAL FIELD Description of environment: Z Obstacles Complete control of the robot behavior Simply implementation No stability problem FIELD of Attractive Forces Target Obstacle difficult environment modeling not easy integration in industrial controller NOT CERTIFIABLE FIELD of Repulsive Forces The Resultant Force Field is applied to the robot kinematic structure and the path is automatically defined The trajectory follow the minimum potential direction Global Optimum THE TARGET IS MOVING WITH THE TCP POSITION 22

23 REACTION STRATEGIES: Online Replanning INDUSTRIAL OBSTACLE AVOIDANCE FRAMEWORK (Input 1) Stl-CAD Model (Output) NEW TARGET TO ROBOT INTERPOLATOR (Input 2) Nominal Trajectory COLLISION AVOIDANCE SYSTEM (Feedback 1) Joint Position (Feedback 2) Obstacle Position Esitmation 23

24 REACTION STRATEGIES: Online Replanning INDUSTRIAL OBSTACLE AVOIDANCE FRAMEWORK (Input 1) Stl-CAD Model (Output) NEW TARGET TO ROBOT INTERPOLATOR Off-line processing (Input 2) Nominal Trajectory COLLISION AVOIDANCE SYSTEM On-line processing (Feedback 1) Joint Position (Feedback 2) Obstacle Position Esitmation 24

25 REACTION STRATEGIES: Online Replanning Off-line processing On-line processing From ENVIRONMENT and PATH model To a GRID of Pass-Through Points Environment repulsive forces GRID Choice of the best point within the grids Human Position SENDING of the NEXT TARGET NODE 25

26 The safe & productive robot working without fences 4. Application Scenarios 26

27 SHOOP FLOOR SCENARIO FEASIBLE SCENARIOS Big robot and extensive trajectories ROBOT BIGGER/EQUAL TO HUMAN OPERATORS Human Operator and robot cooperating in the same task. The robot has to avoid the obstacle CORRECTING the trajectory of the TCP Identification of unforeseen human near the robot is AVAILABLE 27

28 SHOP-FLOOR SCENARIO Humans that INTERACT with the robot Humans NEAR the robot DIFFERENT STRATEGIES & DIFFERENT SENSING SYSTEM 28

29 SHOP-FLOOR SCENARIO (2) Workcell Supervision Humans identification 2D CHEAP CAMERA Humans that INTERACT with the robot Humans NEAR the robot 1. Identification Accurate measure Tracking 3D TOF 2D CHEAPCAMERA 2. Reaction Strategies Merging of different sensors BOG 1. Identification Tracking 2D CHEAP CAMERA 2. Reaction Strategies Trajectory/Velocity Modification INDUSTRIAL COLLISION AVOIDANCE FRAMEWORK Reverse movement WITHDRAW Intrusion detection SLOW DOWN Continuous Velocity Modification FUZZY 29

30 FEASIBLE SCENARIOS DESKTOP APPLICATION Small robot and extensive trajectories ROBOT EQUAL/SMALLER THAN HUMAN OPERATORS Robot interacts with the ARMS of the operators The robot WORKSPACE is LIMITED with respect to mobility of humans Identification of object INNER/OUTER zones around the robot 30

31 DESKTOP SCENARIO Identification IF humans are PRESENT in the application area: It can be achieved using on the shelves sensors Very Simple Supervision Logic Small Workspace Simpler than the solutions developed by the consortiums Already certified..also others have good idea! Human Operators POSITIONS identification Human Hands CONFIGURATION and POSE Accurate measure 3D TOF Local Approach CAPACITIVE Reaction strategies Intrusion detection SAFE ZONES AROUND LINKS Slow Down and Reverse Movement WITHDRAW 31

32 Considerations and Conclusions The safe & productive robot working without fences What has been achieved? Proof of concepts HW and SW Functionality Usability in applications Future activities will be focused on: SAFETY CERTIFICATION OF EXISTING AND NEWLY DEVELOPED SW, HW, SENSORS > one/two years INDUSTRIALIZATION OF THE DEVELOPED SYSTEMS > few years INTEGRATION OF DEVELOPED HW/SW WITH SAFE CERTIFIED CONTROLLERS > some years TRAINING HUMANS OPERATORS 32

33 The European Robot Initiative for Strengthening the Competitiveness of SMEs in Manufacturing Thank you for your kind attention! (