In this work, we investigate a task planning problem for assigning and planning a mobile robot team to jointly perform a kitting application with alternative task locations. To this end, the application is modeled as a Robot Task Scheduling Graph and the planning problem is modeled as a Mixed Integer Linear Program (MILP). We propose a heuristic approach to solve the problem with a practically useful performance in terms of scalability and computation time. The experimental evaluation shows that our heuristic approach is able to find efficient plans, in comparison with both optimal and non-optimal MILP solutions, in a fraction of the planning time.
The efficiency of collaborative mobile robot applications is influenced by the inherent uncertainty introduced by humans' presence and active participation. This uncertainty stems from the dynamic nature of the working environment, various external factors, and human performance variability. The observed makespan of an executed plan will deviate from any deterministic estimate. This raises questions about whether a calculated plan is optimal given uncertainties, potentially risking failure to complete the plan within the estimated timeframe. This research addresses a collaborative task planning problem for a mobile robot serving multiple humans through tasks such as providing parts and fetching assemblies. To account for uncertainties in the durations needed for a single robot and multiple humans to perform different tasks, a probabilistic modeling approach is employed, treating task durations as random variables. The developed task planning algorithm considers the modeled uncertainties while searching for the most efficient plans. The outcome is a set of the best plans, where no plan is better than the other in terms of stochastic dominance. Our proposed methodology offers a systematic framework for making informed decisions regarding selecting a plan from this set, considering the desired risk level specific to the given operational context.
The rapid development of IoT, cloud and fog computing has increased the potential for developing smart services for IoT devices. Such services require not only connectivity and high computing capacity, but also fast response time and throughput of inferencing results. In this paper we present our ongoing work, investigating the potential for implementing smart services in the context of industrial robot applications with focus on analytic inferencing on fog and cloud computing platforms. We review different use cases that we have found in the literature and we divide them into two suggested categories, "distributed deep models" and "distributed interconnected models". We analyze the characteristics of IoT data in industrial robot applications and present two concrete use cases of smart services where inferencing in a fog and a cloud architecture, respectively, is needed. We also reason about important considerations and design decisions for the development process of analytic services.
Industrial mobile robots are increasingly introduced in factories and warehouses. These environments are becoming more dynamic with human co-workers and other uncertainties that may interfere with the robot's actions. To uphold efficient operation, the robots should be able to autonomously plan and replan the order of their tasks. On the other hand, the robot's actions should be predictable in an industrial process. We believe the deployment and operation of robots become more robust if the experts of the industrial processes are able to understand and modify the robot's behaviour. To this end, we present an intuitive novel task modelling formalism, Robot Task Scheduling Graph (RTSG). RTSG provides building blocks for the explicit definition of alternative task sequences in a compact graph format. We present how such a graph is automatically converted to a task planning problem in two different forms, i.e., a Mixed Integer Linear Program (MILP) and a Planning Domain Definition Language specification (PDDL). Converted RTSG models of a mobile kitting application are used to experimentally compare the performance of one MILP planner and two PDDL planners. Besides providing this comparison, the experiments confirm the equivalence of the converted MILP and PDDL problem formulations. Finally, a simulation experiment verifies the assumed correlation between a cost model, based on path lengths, and the makespan.
In the original article, Listings 1 and 2 were not included during the typesetting process and were overlooked during production. The missing listings appear below.
Modern industrial robots are increasingly deployed in dynamic environments, where unpredictable events are expected to impact the robot's operation. Under these conditions, runtime task replanning is required to avoid failures and unnecessary stops, while keeping up productivity. Task replanning is a long-sighted complement to path replanning, which is mostly concerned with avoiding unexpected obstacles that can lead to potentially unsafe situations. This paper focuses on task replanning as a way to dynamically adjust the robot behaviour to the continuously evolving environment in which it is deployed. Analogously to probabilistic roadmaps used in path planning, we propose the concept of Task roadmaps as a method to replan tasks by leveraging an offline generated search space. A graph-based model of the robot application is converted to a task scheduling problem to be solved by a proposed Branch and Bound (B&B) approach and two benchmark approaches: Mixed Integer Linear Programming (MILP) and Planning Domain Definition Language (PDDL). The B&B approach is proposed to compute the task roadmap, which is then reused to replan for unforeseeable events. The optimality and efficiency of this replanning approach are demonstrated in a simulation-based experiment with a mobile manipulator in a kitting application. In this study, the proposed B&B Task Roadmap replanning approach is significantly faster than a MILP solver and a PDDL based planner.