Content. Robotik: Möglichkeiten, Trends und Visionen. Introduction. Robotics the challenges and technology drivers. Robot Examples

Transcription

1 Robotik: Möglichkeiten, Trends und Visionen Roland Siegwart, ETH Zurich Helbling-Abendseminar 18. März 2015, Swissôtel Zürich Roland Siegwart Content Introduction Robotics the challenges and technology drivers Robot Examples Rolling Running Flying Household Opportunities / Markets Roland Siegwart

2 Research Mission and Dedication to create intelligent robots and systems that are able of operating autonomously in complex and dynamic environments. Research Foci novel robot concepts that are best adapted for acting on the ground, in the air, or in the water. Navigation concepts for autonomous operation in challenging environments. Roland Siegwart ASL ETH Zurich Micro Air Vehicles Walking and Running Quadruped Robots Service Robots Autonomous Robots/Cars for Inner City Environments Inspection Robots Space Robots for Planetary Exploration Swimming Robots Prof. Dr. Roland Siegwart 15

3 Autonomous mobile robot the see-think-act cycle knowledge, data base mission commands Localization Map Building position global map Cognition Path Planning environment model local map path Perception Information Extraction raw data Sensing see-think-act Path Execution actuator commands Acting Motion Control Real World Environment Roland Siegwart Robotics the challenges and technology drivers The challenges Seeing, feeling and understanding the world Dealing with uncertain and only partially available information The technology drivers Sensor Technologies Computational power Actuator technologies Technology evolutions enable robotics revolutions Laser time-of-flight sensors Cameras and IMUs combined with required calculation power Torque controlled motors New materials Roland Siegwart

4 Seeing the world a world full of uncertainties Reasoning about a situation Dealing with the real world Cognitive systems have to analyze and interpret situations based on uncertain and only partially available information The need ways to learn functional and contextual information (models / semantics / understanding) Probabilistic Reasoning Roland Siegwart Seeing the world more than appearance Perception and models ( understanding ) are strongly linked What is the difference in brightness? Roland Siegwart

5 Seeing the world more than appearance Perception and models ( understanding ) are strongly linked What is the difference in brightness? Roland Siegwart Feeling the world torque controlled actuation Tactility, key for controlling the real world Courtesy of Albu-Schaeffer & Hirzinger, DLR, Germany It takes us around 14 years to learn holding a glass with an optimal force

6 Understanding the world Bayesian Reasoning Reasoning in the presence of uncertainties and incomplete information Combining preliminary information and models with learning from experimental data Picture Courtesy of Bessiere, INRIA Grenoble, France Roland Siegwart Localization where am I? SEE: The robot queries its sensors finds itself next to a pillar ACT: Robot moves forward motion estimated by wheel encoders accumulation of uncertainty SEE: The robot queries its sensors again finds itself next to a pillar Belief update (information fusion) Roland Siegwart

7 Learning the world real-time adaptation Courtesy of Aude Billard EPFL Roland Siegwart 30 Understanding the world Fusing & Compressing Information Places / Situations A specific room, a meeting situation, Servicing / Reasoning Interaction Navigation Objects Doors, Humans, Coke bottle, car, Features Lines, Contours, Colors, Phonemes, Raw Data Vision, Laser, Sound, Smell, Functional / Contextual Relationships of Objects imposed learned spatial / temporal/semantic Models / Semantics imposed learned Models imposed learned Roland Siegwart

8 Understanding the world Humans are unbeatable in taking decisions in complex situations Technology is better in simple but fast decisions (ABS, ESP, ) Complexity of Services Tactile Mobile Manipulation Position Based Mobile Manipulation Advanced Dialog Autonomous Navigation Actions from simple motion to complex interaction Robotics Roadmap Static Structured, 2D Static Toys AGVs Household assistant Industrial services Tour-Guides Household universal All-terrain navigation Static Unstructured, 3D Agriculture robots Search and Rescue Autonomous car freeway Dynamic Structured, 2D Autonomous car urban Dynamic Unstructured, Dynamic3D Semantics dynamic Environment - from static 2D grid maps to 3D cognitive maps

9 Rolling Robots some examples Roland Siegwart Laser-based outdoor navigation Roland Siegwart

10 Roland Siegwart V-Charge Automonous driving using close-to-market sensors Expensive and complex Roland Siegwart

11 V-Charge Autonomous driving using close-to-market sensors Wheel encoders ultrasound cameras Roland Siegwart V-Charge a typical scenario Roland Siegwart

12 V-Charge Vision and Results Roland Siegwart Running Robots Roland Siegwart

13 Efficient Walking and Running what nature evolved (Extreme Jumpy Dog) Roland Siegwart Spot hydraulic quadruped Roland Siegwart

14 Efficient Walking and Running serial elastic actuation Roland Siegwart StarlETH Leg Design for Dynamic Walking High fidelity DCmotors with harmonic drives Leaf springs Planar 3-DoF guiding unit Chain drive reduces leg inertia High resolution encoders to register forces damper Modular foot design Nonlinear spring for knee Roland Siegwart

15 StarlETH agile, efficiency and robust precise torque control during stance fast task space position control during swing virtual model controller for ground contact autonomous gait discovery by stochastic optimization Roland Siegwart Swimming Robots some examples Roland Siegwart

16 Sepios The calmar robot Roland Siegwart Environmental monitoring Monitoring (e.g. algaes) in lakes LizhbETH Roland Siegwart

17 Flying Robots some examples Roland Siegwart UAV (Unmanned Aerial Vehicles) flight concepts Helicopters: < 20 minutes Highly dynamic and agility Fixed Wing Airplanes: > some hours; continuous flights possible Non-holonomic constraints Blimp: lighter-than-air > some hours (dependent on wind conditions); Sensitive to wind Large size (dependent on payload) Flapping wings < 20 minutes; gliding mode possible Non-holonomic constraints Very complex mechanics Festo BionicOpter Roland Siegwart

18 UAV potential applications Search and rescue, surveillance Industrial inspection Agriculture, mining and construction Next generation satellites Roland Siegwart UAV requirements Appropriate flight concept Power autonomy Agility Robustness Navigation with on-board sensing and processing Robustness against communication and GPS loss home button Simple and intuitive operation Stable on hands-off Collision avoidance and localization / SLAM Roland Siegwart

19 UAV Vision only navigation Swarm of small helicopters Vision only navigation (one camera, GPS denied) Fully autonomous with on-board computing Feature-based visual SLAM robust against lighting changes and large scale changes Proto 1 Proto 2 Proto 3 Roland Siegwart A Synchronized Visual-Inertial Sensor System with FPGA Pre-Processing for Accurate Real-Time Slam Roland Siegwart

20 Keyframe VIO with Online Extrinsics Estimation Handheld around ETH MSCKF: visual-inertial stochastic cloning slidingwindow filter (Mourikis et al., 2009). Roland Siegwart UAV collision avoidance and path planning Real time 3D mapping (on-board) optimal path planning considering localization uncertainties Proto 1 Proto 2 Proto 3 Roland Siegwart

21 UAV facade scanning and 3D reconstruction Enhanced teleoperation or autonomous operation Visual-inertial localization for optimal 3D reconstruction Proto 1 Proto 2 Proto 3 Roland Siegwart UAV 3D mapping in mines Vision-based localization and SLAM Laser-based 3D mapping Roland Siegwart

22 overview an omnidirectional, spherical aircraft Actuation Unit Skye Bus Electronics and Camera Total Weight Actuation Units (4x) Electronics and Power Hull Pressure Elements kg kg kg kg Ca Kg Buoyancy Ca. 10 kg Diameter Ca. 2.7 m Volume Ca. 10 m 3 Roland Siegwart Solar Airplane design methodology for continuous flights Based on Mass & Power Balance Need for precise scaling laws (mass models) Airplane Parts Solar cells Battery Airframe Total mass Aerodynamic & Conditions Power for level Flight Roland Siegwart

23 Solar Airplane Optimization Design space at 38 N, June 21 st Battery mass [kg] Fixed Aspect Ratio: Excess Time [h] Wingspan [m] Flat optimum at wingspan 11.5 m Chosen AtlantikSolar configuration: Wingspan 5.65 m Battery mass 2.9 kg Structural weight Predicted: g Effective: g Prediction [Noth 08]: g Roland Siegwart Solar powered fixed wing airplanes: Long duration / continuous flights sensesoar Wingspan: 3 m Wing area: m2 Peak Solar power 140 W Power Consumption 50 W Masses: Overall: 3.72 kg Batteries: 1.89 kg Nominal Speed 10 m/s Sensors Air speed IMU GPS Camera IR camera AtlantikSolar Wingspan: 5.64 m Solar area: 1.5 m2 Peak Solar power 280 W Power Consumption 40 W Masses: Overall: 6.2 kg Batteries: 2.9 kg Nominal Speed 10 m/s Sensors Air speed IMU GPS Camera Roland Siegwart

24 crossing the Atlantic Boston Lisbon Around 6 days without tail wind Roland Siegwart Solar Airplane visual navigation Visual-inertial sensor with multiple cameras Integrated thermal vision Robust state estimation and flight control Autonomous planning for complete inspection Long endurance solar powered fight Roland Siegwart

25 ARMAR-III Karlsruhe Institute of Technology PR2 Willow Garage Household Robots some examples Roland Siegwart ARMAR-III in the kitchen Courtesy of Karlsruhe Institute of Technology

26 Perceiving and Handling of Objects the PR2-Robot from Willow Garage Fold towels Courtesy of Clean-up with cart Roland Siegwart Opportunities / Markets Industrial transportation Cleaning Medical robotics Entertainment / edutainment Autonomous Cars Office logistics Industrial services Surveillance and rescue Construction and mining Agriculture The coffee servant Nesspresso / Bluebotics, Switzerland Health and elderly care Personal / services robots Roland Siegwart

27 KIVA Robots the other warehouse automation Excerpt from History Channel, 2008 Acquired by Amazon in 2012 Key Industrial Collaborations and Partnerships Roland Siegwart

28 ASL Team Roland Siegwart Switzerland, the Silicon Valley of Robotics Chris Anderson CEO of 3DRobotics WIRED, editor-in-chief until 2012 Roland Siegwart