Follow me

Tuesday, October 2, 2012

Real-Time Safety for Human – Robot Interaction


Dana Kulic, Elisabeth A. Croft
International Conference of Industrial Robotics, 2005
Summary
                  Typical robotic industrial applications see robots segregated in order to ensure safety, this is not acceptable in more and more flexible environments which require interaction with operators, while ensuring safety conditions. In order to ensure safety a complete system must then incorporate the following features: safe mechanical design, human friendly interfaces, safe planning and control strategies; the authors focus on the third aspect. There in the literature mainly three methods to mitigate risk: redesign to eliminate hazard, control the hazard through electronic or physical safeguards and warn the operator during operation or by training. While the last has been proven to not be effective, specially in presence of untrained users, the first one appears to be the case in which the aim is reducing the results of a possible impact and not the impact itself.
The key issue then is to understand when safety is threatened, this can be done using tactile and force sensors to identify unplanned contact or by considering every object in the environment as an obstacle and use real-time obstacle avoidance strategy. The proposed algorithm introduces a danger index, if the level of danger is low the plan can proceed, otherwise there will be a corrective decision, this safety system performs then in one-step-ahead condition, moving towards the location with the lowest danger.
The danger index appear to be the multiplication of distance factor, velocity factor and inertia factor, the multiplication ensures no active evasive action which would cause the evasion of the robot also in safe conditions.
The Distance factor is obtained considering a scaling factor kD and it is build so that when the distance between the critical point and the object is large (greater than Dmax) then the factor would go to 0. In a similar fashion the velocity factory is obtained. The inertia factor appears to be the ratio between the effective inertia at the critical point and the maximum safe value of the robot inertia.
·       Model
A one dimensional example is performed showing the robot’s behavior when there are not obstacle, one obstacle and two obstacles on the two robot’s sides. It is shown that in the case of two obstacles, the robot is capable of avoiding the obstacles and reaches a stable point with no oscillation.
The full algorithm (regarding all the robot arm) results in a local sequential planner, backtracking cannot be considered since it’s the case of real-time motion. In this case the damping force is obtained as: Fd=-Bq’ where q’ is the measured joint velocity. To calculate distances a set of spheres representation is used, being a ceservative method ensuring safety.
Parameter selection is obtained through physical characteristics of the robotic arm, Dmin can be estimated based on the maximum robot deceleration and velocity and Vmax can be estimated on indices of injury such as Gadd Severity Index or Head Injury Criterion.
Dmax is based on the physical size of the robot and Vmin must be smaller than 0 in order to ensure stability in the system.
Ranges for distances and velocity boundaries should be large so that there is time to react before the danger is imminent, for the velocities this ensures reduction of the effective damping.
Key Concepts
Safety, danger index
Key Results
Simulation has confirmed expected behavior, showing no oscillation in case of two obstacle in the robots proximity. The set of spheres ensures for the experiment a conservative system, so that highly accurate position sensing is not required. The method can be used for redundant and non-redundant manipulators, generating robot motion by minimizing the danger index.

No comments: