Robot Collision Detection Using Momentum Observer Compensation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing collision detection methods for robots, particularly humanoid robots, rely on acceleration measurements that introduce high noise and require additional sensors, making them less effective and more complex.

Innovation Solution

A method that uses a kinematic chain structure with sensors for measuring force/torque and proprioceptive data to estimate generalized external forces and detect collisions without acceleration measurements, employing a momentum observer and compensation for rigid body dynamics and gravity effects to accurately identify collision locations and forces.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If acceleration measurements are used for collision detection, then collision detection capability is improved, but measurement noise increases significantly

Engineering Contradiction:
Improvecollision detection capabilityVSAvoidmeasurement noise
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent extracts and removes the acceleration measurement component from the collision detection system. Instead of using acceleration measurements, the invention relies solely on force/torque sensor data and momentum observer techniques, thereby eliminating the source of high measurement noise while maintaining collision detection capability

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces the mechanical acceleration measurement approach with a computational momentum-based observer system. The momentum observer uses force/torque sensor measurements and dynamic model information to estimate collision forces, substituting direct mechanical acceleration sensing with an indirect computational estimation method that avoids noise

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If additional sensors are added for collision detection, then measurement accuracy is improved, but device complexity increases

Engineering Contradiction:
Improvecollision detection accuracyVSAvoidsensor quantity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent makes the force/torque sensors serve multiple functions: they are used for both manipulation control and collision detection. By properly utilizing the existing force/torque sensor data through momentum observer techniques and dynamic model compensation, the system achieves collision detection capability without requiring additional dedicated sensors

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses its own existing sensor infrastructure (force/torque sensors) to perform collision detection. The momentum observer leverages the robot's dynamic model and existing force/torque measurements to self-generate collision detection information, making the system self-sufficient without external additional sensing

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS11370117B2Collision handling by a robot
Publication Date: 2022.06.28 FR ADMINISTRATION GMBH
  • US11370117B2 patent drawing
  • US11370117B2 patent drawing
  • US11370117B2 patent drawing

AI summary

The invention relates to a method of collision handling for a robot with a kinematic chain structure comprising at least one kinematic chain, wherein the kinematic chain structure includes: a base, links, joints connecting the links, actuators and at least one end-effector, a sensor Sdistal.i in the most distal link of at least one of the kinematic chains for measuring/estimating force/torque, and sensors Si for measuring/estimating proprioceptive data, wherein the sensors Si are arbitrarily positioned along the kinematic chain structure, the method including: providing a model describing the dynamics of the robot; measuring and/or estimating with sensor Sdistal.i force/torque Fext,S.distal.i in the most distal link of at least one of the kinematic chains; measuring and/or estimating with the sensors Si proprioceptive data: base and robot generalized coordinates q(t) and their time derivative {dot over (q)}(t), generalized joint motor forces τm, external forces FS, a base orientation φB(t) and a base velocity {dot over (x)}(t)B; generating an estimate {circumflex over (τ)}∈ of the generalized external forces τext with a momentum observer based on at least one of the proprioceptive data and the model; generating an estimate {umlaut over ({circumflex over (q)})}(t) of a second derivative of base and robot generalized coordinates {umlaut over (q)}(t), based on {circumflex over (τ)}∈ and τm; estimating a Cartesian acceleration {umlaut over ({circumflex over (x)})}D of point D on the kinematic chain structure based on {umlaut over ({circumflex over (q)})}(t); compensating the external forces FD for rigid body dynamics effects based on {umlaut over ({circumflex over (x)})}D and for gravity effects to obtain an estimated external wrench {circumflex over (F)}ext,S.i; compensating {circumflex over (τ)}∈ for the Jacobian JS.distal.iT transformed Fext,S.distal.i to obtain an estimation {circumflex over (τ)}ext,col of generalized joint forces originating from unexpected collisions; detecting a collision based on given thresholds τthresh and FS.i,thresh if {circumflex over (τ)}ext,col>τthresh and/or if {circumflex over (F)}ext,S.i>FS.i,thresh.