Context-dependent Compensation Scheme to Reduce Robot Trajectory Execution Errors
Background:
A robot’s repeatability measures how precisely the robot can return to a previously visited (or taught) workspace configuration (e.g., end-effector position and orientation) under the same loading configuration. The accuracy of a robot measures how accurately a robot can reach a specified workspace configuration. This requires inverse kinematics calculations to compute the joint angles and using a feedback controller to achieve the desired joint values. The kinematic errors (e.g., link dimensions, parallelism, orthogonality errors, etc.), joint errors (e.g., sensor errors, etc.), and the non-kinematic errors (e.g., link stiffness, gear backlash, etc.) of the robot result in deviation from the desired end-effector configuration.Â
Currently, automatically generated trajectories cannot be directly used on tasks that require high execution accuracies due to errors accused by inaccuracies in the robot model, actuator errors, and controller limitations. These trajectories often need manual refinement. This is not economically viable on low production volume applications. Unfortunately, execution errors are dependent on the nature of the trajectory and end-effector loads, and therefore devising a general-purpose automated compensation scheme for reducing trajectory errors is not possible.
Our laboratory’s focus:
To mitigate these errors, we are developing a method for analyzing the given trajectory, executing an exploratory physical run for a small portion of the given trajectory, and learning a compensation scheme based on the measured data. The learned compensation scheme is context-dependent and can be used to reduce the execution error.
Self-Supervised Learning of Process Parameters
Background:
Robots are traditionally used in assembly, machine tending, and material handling tasks. There is increasing interest in using robots for processing applications. In these applications, the robot is required to make changes to the part being processed. Robotic surface finishing tasks such as cleaning, polishing, sanding, machining, and painting are examples of processing applications. Physics-based models are not known a priori for many processing applications. This is often the case when a new material, part, or tool is under consideration. This problem also arises when there is significant uncertainty in the state of the workpiece or the tool.
A large number of physical experiments need to be conducted to build a complete physics-based model. This can be done for large volume production. Building a complete model first and then using it for parameter optimization is not practical for low volume production. Moreover, the complete model might not be reused due to the nonrepetitive nature of tasks in small volume manufacturing.
Our laboratory’s focus:
Our interest is in developing a self-supervised learning method that can identify the optimal operation and trajectory parameters with a small number of experiments. We are developing a method that optimizes the task objective and meets the task performance constraints. The algorithm makes decisions based on the uncertainty in the surrogate models of task performances. The method intelligently samples the parameter space to select a point for evaluation from the sampled set by determining its probability to be optimum within the set. The iterative method rapidly converges to the optimal point with a small number of experiments. We have validated our approach through physical experiments of robotic scrubbing and sanding.
Tool Path Planning for Ensuring Correct Tool Part Contacts
Background:
Tool-Path planning is the foundation for automating many manufacturing processes. Robotic manipulators are increasingly being considered to automate tasks that require complex tool motions. Robotic manipulators provide extra degrees of freedom and are more flexible than traditional automation technologies. However, a tool-path needs to be planned and given to the manipulator trajectory generator as input. The traditional tool-path planning considers the tool to have one contact point, known as the Tool Center Point (TCP). This underutilizes the available flexibility of the manipulator. To make use of the manipulator’s flexibility, multiple contact points or multiple TCP candidates can be considered. These tool contact considerations make the tool path planning problem complex and computationally challenging.
Our laboratory’s focus:
We tackle these problems with a novel tool path planning algorithm. Our algorithm incorporates the multiple tool contact points consideration during tool-path planning in an efficient manner to generate a high-quality tool-path in a reasonable amount of run time. The numerous possible tool-paths generated due to the multiple TCP candidates are encoded in a directed graph. The efficient use of collision checking on the tool-path generated by a graph search algorithm is used to converge to the high-quality tool-path quickly.
Alert Generation for Human-Robot Teams
Background:
Multi-robot systems can be particularly useful for dangerous missions in complex environments, e.g., disaster response, and reduce the risk to humans.
We anticipate that humans will be in supervisory roles to make decisions in critical situations, and manage resources. At a high-level, humans will ensure that robots are continuously working on tasks that align with the mission objectives as the mission progresses. In a mission with considerable uncertainty due to intermittent communications, degraded information flow, and failures, humans need to assess both the current and probable future states to make superior decisions.
Our laboratory’s focus:
We are developing a forward simulation-based alert system that proactively notifies the human supervisor of possible undesirable events, which assists humans to re-task agents accordingly. We are also developing methods for speeding up mission simulations to ensure real-time alert generation, essential in time-critical missions. We have conducted human subject studies to verify that these alerts do improve the decision making performance of human supervisors.