Message Expiration-Based Distributed Multi-Robot Task Management

Since the early 1990s, the problem of task allocation in multi-robot systems has received significant, and increasing interest in the research community [1]. As researchers design, build and use cooperative multi-robot systems, they invariably encounter the question: ”Which robot should execute which task?” This question must be answered, even for relatively simple multi-robot systems, and the importance of task allocation grows with the complexity, size, and capability of the system under study. Even in the simplest case of homogeneous robots with fixed, identical roles, intelligent allocation of tasks is required for good system performance, if only to minimize physical interference.

Recent advances in communication and computation have given rise to distributed control of multiagent systems which, compared to conventional centralized control, provide increased efficiency, performance, and scalability as well as robustness due to its ability to adjust to agent failures or dynamically changing environments [2, 3, 4]. Inspired by these appealing properties of distributed control, we propose a distributed and online solution to the multiagent dynamic task allocation problem, where the assignment of robots to tasks may need to be continuously adjusted depending on changes in the task environment.

This paper focuses on multi-robot task management to achieve high performance and efficiency in sample foraging tasks with multiple homogeneous (swarm) robots. Additionally, we are dealing with a distributed multi-robot system instead of a centralized multi-robot system to improve the scalability of the proposed method.

Most multi-robot task allocation algorithms focus on the efficiency of task accomplishment based on strict assumptions. The assumptions include that the energy level of the robot will not be a problem or that the communication of the robots is always stable. However, in practice, there are many constraints on the robots, especially for the swarm that they are designed in a small shape, which leads to poor communication signals and limited battery [5]. Similarly, swarm robots cannot have a sufficient battery to finish multiple tasks without recharging due to physical constraints. With an excellent recharging algorithm, the swarm robots can work on their jobs automatically. On the other hand, heterogeneity in the robot’s capability such as their max speed, energy level, variations on functional intelligence, etc. can also impact the task performance [6].

Yang et al. [7, 8] introduced a novel framework for cooperation between heterogeneous multi-robot systems for dynamic task assignments and automated planning. It combines robot perception, communication, planning, and execution in MRS, which considers individual robots’ needs and action plans and emphasizes the complex relationships created through communication between the robots. Specifically, they proposed Robot’s Needs Hierarchy to model the robot’s motivation and offer priority-based task management. In [9], the authors presented a distributed energy management task allocation strategy for multi-robot systems involved in goods transportation, considering energy management in the task allocation. Authors in [10] presented a control barrier function-based framework for long-term autonomy of multi-robot systems with limited charging resources.

Motivated by these two demands on energy and communication for swarm robots, we propose a novel distributed algorithm that can handle the disconnectivity and the recharging problem and achieve continuous performance in repetitive tasks. Specifically, this paper presents a multi-robot task management algorithm integrative of energy-awareness and connectivity-awareness focused on distributed tasks by introducing a new expiration-based validation mechanism to avoid conflicts in task allocation.

Let us denote $R=\{R_{1},R_{2},...,R_{n}\}$ as the multi-robot systems including $n$ Robots. For each robot $R_{i}$, it will have its neighbors, such as $N_{i}=\{R_{1},R_{3},R_{5}\}$. What is more, for each robot $R_{i}$, it will have its own task schedule including other robots information, denoted as $S_{i}=\{S_{i}^{1},S_{i}^{2},...,S_{i}^{n}\}$.

The problem is to synchronize the task schedule over each robot with the following constraints: the connectivity may vary due to the sensing range. Because of the unstable connectivity, one robot may not be able to synchronize with all other robots.

The proposed task management algorithm involves three main components: 1) task receipt handler; 2) robot task scheduler; 3) robot task controller.

To verify the effectiveness of the proposed approach, we used the simulator version of the Robotarium platform [11], which is a multi-robot testbed. In our simulations, the following parameters are used.

• •

  N=5 number of robots.

• •

  T=5 fixed treasure locations.

• •

  R=2 fixed recharger stations at the top right and top left of the workspace.

• •

  C=1 one fixed collection point at the bottom left of the workspace.

• •

  I=10000 iterations for simulation

• •

  Starting position of the robots = random within the workspace.

• •

  Distance clearance threshold = 0.1m (to decide if a robot has reached a location or not).

• •

  All distances are expressed in the Euclidean norm, and time is expressed as iterations.

• •

  Disable the collision avoidance and barrier function in Robotarium.

Fig. 1 shows an example of the simulation environment. The line between two robots represents they are connected. The black dots represent treasure, recharge station, and collection point.

Assumption: Every task location has an unlimited amount of treasure available. When a robot is at the task location, it is assumed that it picked up the treasure (if it does not have one already). When a robot reaches a collection point, then it is assumed that it dropped the treasure (if it already had one with it).

Definition of Task completion: A task is said to be complete when a robot reaches any of the task locations, and it had then dropped that treasure at the collection point.

Energy model: A robot will lose energy based on the following energy consumption model [12]:

where $\alpha$ represents the coefficient for the static energy consumption due to onboard computing and sensors of the robot for example, $\beta$ represents the coefficient for the dynamic power consumption by the robot due to movement (e.g, velocity-based motor power to move to a specific distance), and $\gamma$ represents the coefficient for the energy spent in picking up the treasure (e.g., manipulation effort).

Similarly, a robot will gain energy if it is at the recharging station with the following model:

where $\delta$ represents the rate at which the battery can be recharged. We use the following values: $\alpha=0.01,\beta=0.1,\gamma=0.02,\delta=0.1$.

We proposed a novel expiration-validation-based approach for multi-robot task management of persistent tasks without conflicts on an energy-constrained swarm robotics system. The proposed expiration mechanism improved the performance when the sensing range of the robots was small.