Human-Robot Task and Actions planning in HRC applications

Human-Robot Collaborative (HRC) workcells are a novel approach in the industry, facilitating seamless interaction and collaboration between humans and robots.

However, in HRC manufacturing lines, it is imperative to encompass modules able to perform task planning, adapt the behavior of resources to dynamic events and evaluate the metrics of the assembly procedure, such as operation time and quality of performance.

Likewise, SHERLOCK’s Task and Action Planning Software module’s goal is to incorporate flexibility in the product assembly by planning according to specific optimization needs.

The planning result consists of assigning the operator and the collaborative robot with tasks needed to complete the product to resources capable of executing them in such a way as to optimally satisfy the required KPIs. For example, the total distance covered by the operator, the total time needed for the product execution or even ergonomic and safety thresholds set.

TAPS planning output and execution schedule

The planning result is generated based upon a stochastic algorithm and comprises of multiple heuristic and search functions that aim to improve specific metrics of the assembly, depending on the criteria configuration used.

For more accurate calculation of said metrics, a digital replica of the workspace is being utilized, to simulate the different possible alternative product assemblies that are being examined and compared.

The developed solution needs to be able to adapt to the inevitably unpredictable behavior of an operator. In case of deviation from the original plan or of inadequate satisfaction of KPIs, online rescheduling of the remaining tasks is supported, so that the remaining tasks can be planned again, aiming to satisfy the same criteria configuration but obviously using less processing power as the process needs to be done instantaneously and reschedule re-commence the manufacturing procedure.

Additionally, a necessary integration with the orchestrator module was implemented. Each high-level task-resource assignment contains a corresponding (low-level) diagram of actions, for example robot movements or operator instructions. Combined, the assignments of the tasks, and the actions they are individually comprised of, create an execution schedule.

The execution schedule is a corresponding directed graph of conditionally connected nodes of assignments, as it includes logical operators and it is event-driven (e.g., in case of failed actions, different paths are followed)

Every time an assembly plan is generated, the time-ordered assignments (generated from the planning procedure) are being converted into a schedule form that the orchestrator utilizes for monitoring and orchestration of the execution.

Finally, a way to quickly configure planning and execution data of new product was implemented. Goal of this implementation was to be able to smoothly and easily create new products, even without the attendance of the developers of the SHERLOCK modules. This guided variant creation leverages on the preexisting modelling of the project and streamlines the creation of product plans that follow a generic pattern either via a csv or by a dedicated UI.

Assembly tasks chart