Classifying Environmental Features From Local Observations of Emergent Swarm Behavior

Robots in a swarm are programmed with individual behaviors but then interactions with the environment and other robots produce more complex, emergent swarm behaviors. One discriminating feature of the emergent behavior is the local distribution of robots in any given region. In this work, we show how local observations of the robot distribution can be correlated to the environment being explored and hence the location of openings or obstructions can be inferred. The correlation is achieved here with a simple, single-layer neural network that generates physically intuitive weights and provides a degree of robustness by allowing for variation in the environment and number of robots in the swarm. The robots are simulated assuming random motion with no communication, a minimalist model in robot sophistication, to explore the viability of cooperative sensing. We culminate our work with a demonstration of how the local distribution of robots in an unknown, office-like environment can be used to locate unobstructed exits.      

The pros and cons of flocking in the long-range “migration” of mobile robot swarms

In this paper, we study how flocking affects the accuracy and speed of individuals in long-range “migration”. Specifically, we extend a behavior that can generate self-organized flocking in a swarm of robots to follow a homing direction sensed through the magnetic field of the Earth and evaluate how the final points reached by the flock are scattered in space and how the speed of the flock is affected. We propose that four factors influence the performance of migration, in the proposed behavior, namely: (1) averaging through heading alignment behavior, (2) disturbances caused by proximal control behavior, (3) noise in sensing the homing direction, and (4) differences in the characteristics of the individuals. Systematic experiments are conducted to evaluate the effects of these factors using both physical and simulated robots. The results show that although flocking reduces the speed of an individual, it increases the accuracy of “migration” for flocks that are larger than a certain size.     

Swarm robotics search & rescue: A novel artificial intelligence-inspired optimization approach

In this paper, a novel heuristic algorithm is proposed to solve continuous non-linear optimization problems. The presented algorithm is a collective global search inspired by the swarm artificial intelligent of coordinated robots. Cooperative recognition and sensing by a swarm of mobile robots have been fundamental inspirations for development of Swarm Robotics Search & Rescue (SRSR). Swarm robotics is an approach with the aim of coordinating multi-robot systems which consist of numbers of mostly uniform simple physical robots. The ultimate aim is to emerge an eligible cooperative behavior either from interactions of autonomous robots with the environment or their mutual interactions between each other. In this algorithm, robots which represent initial solutions in SRSR terminology have a sense of environment to detect victim in a search & rescue mission at a disaster site. In fact, victim’s location refers to global best solution in SRSR algorithm. The individual with the highest rank in the swarm is called master and remaining robots will play role of slaves. However, this leadership and master position can be transitioned from one robot to another one during mission. Having the supervision of master robot accompanied with abilities of slave robots for sensing the environment, this collaborative search assists the swarm to rapidly find the location of victim and subsequently a successful mission. In order to validate effectiveness and optimality of proposed algorithm, it has been applied on several standard benchmark functions and a practical electric power system problem in several real size cases. Finally, simulation results have been compared with those of some well-known algorithms. Comparison of results demonstrates superiority of presented algorithm in terms of quality solutions and convergence speed.     

Trophallaxis within a robotic swarm: bio-inspired communication among robots in a swarm

This article presents a bio-inspired communication strategy for large-scale robotic swarms. The strategy is based purely on robot-to-robot interactions without any central unit of communication. Thus, the emerging swarm regulates itself in a purely self-organized way. The strategy is biologically inspired by the trophallactic behavior (mouth-to-mouth feedings) performed by social insects. We show how this strategy can be used in a collective foraging scenario and how the efficiency of this strategy can be shaped by evolutionary computation. Although the algorithm works stable enough that it can be easily parameterized by hand, we found that artificial evolution could further increase the efficiency of the swarm’s behavior. We investigated the suggested communication strategy by simulation of robotic swarms in several arena scenarios and studied the properties of some of the emergent collective decisions made by the robots. We found that our control algorithm led to a nonlinear, but graduated path selection of the emerging trail of loaded robots. They favored the shortest path, but not all robots converged to this trail, except in arena setups with extreme differences in the length of the two possible paths. Finally, we demonstrate how the flexibility of collective decisions that arise through this new strategy can be used in changing environments. We furthermore show the importance of a negative feedback in an environment with changing foraging targets. Such feedback loops allow outdated information to decay over time. We found that task efficiency is constrained by a lower and an upper boundary concerning the strength of this negative feedback.     

An improved model for swarm robotics in surveillance based on cellular automata and repulsive pheromone with discrete diffusion

ABSTRACT

An inverted ant cellular automata model called IACA-DI is proposed for the coordination of a swarm of robots performing the surveillance task. The swarm communicate indirectly through the repulsive pheromone, which is available as neighborhood information. The pheromone is deposited at each time step by each robot over its neighborhood. The new model started from a previous one named IACA. However, a discrete modeling of the pheromone diffusion is used in IACA-DI returning a sparser distribution of the robots over the environment. Next movement decisions are based on stochastic cellular automata rules that use the pheromone levels in the neighborhood to perform a probabilistic draw. While in IACA all the neighborhood cells participate in this draw, just a subgroup of them participate in the IACA-DI. It is formed by elite cells ? those with the lowest pheromone levels - and some random selected ones. Besides, the cell that keeps the current robot’s direction receives an increment in its probability to be chosen, giving an inertial tendency to the robot motion. Simple simulations were performed enabling to refine parameters and to choose the better strategies. After this refinement, the resultant model was implemented in the simulation platform Webots? aiming to evaluate IACA-DI with real-world robotic architecture in more realistic scenarios.

IACA-DI is a new model for the coordination of robot swarms performing the surveillance task. It is based on cellular automata modeling and the swarm communicate indirectly through the repulsive pheromone deposited by the robots in the environment cells. Letters (a) and (b) show two snapshots from a simulation of a 3-robots swarm performing the surveillance task. The robots start at random positions in an environment composed by 7 rooms in (a). Thus, based on the IACA-DI decisions, they start to make steps to explore the environment aiming to cover all the rooms in a short interval of time. The trace of each robot after 100 time steps is shown in (b) by representing each individual trajectory with a different color. The behavior of each robot is managed by the IACA-DI model, which can be represented by the FSM with 4 states in (c). Different strategies and formulations were investigated for the two major states ‘next position decision’ and ‘pheromone deposition’. The resultant IACA-DI model is analyzed here using simulations performed in Webots? platform -as the snapshots shown in (a) and (b) -with real-world robotic architectures.     

Adaptive data-driven error detection in swarm robotics with statistical classifiers

Swarm robotics is an example of a complex system with interactions among distributed autonomous robots as well with the environment. Within the swarm there is no centralised control, behaviour emerges from interactions between agents within the swarm. Agents within the swarm exhibit time varying behaviour in dynamic environments, and are subject to a variety of possible anomalies. The focus within our work is on specific faults in individual robots that can affect the global performance of the robotic swarm. We argue that classical approaches for achieving tolerance through implicit redundancy is insufficient in some cases and additional measures should be explored. Our contribution is to demonstrate that tolerance through explicit detection with statistical techniques works well and is suitable due to its lightweight computation.     

Compound behaviors in pheromone robotics

We are pursuing techniques for coordinating the actions of large numbers of small-scale robots to achieve useful large-scale results in surveillance, reconnaissance, hazard detection, and path finding. Using the biologically inspired notion of “virtual pheromone” messaging, we describe how many coordinated activities can be accomplished without centralized control. By virtue of this simple messaging scheme, a robot swarm can become a distributed computing mesh embedded within the environment, while simultaneously acting as a physical embodiment of the user interface. We further describe a set of logical primitives for controlling the flow of virtual pheromone messages throughout the robot swarm. These enable the design of complex group behaviors mediated by messages exchanged between neighboring robots.     

Machine learning for human movement understanding

Main purpose of this project is to develop fundamental technology for assist robots to recover and maintain human motor skill and to extend scope of human activity. Our goal is to provide a system that adapts to its user’s personal behavior patterns in real-time. We aim to develop a continuous collaboration system between the assist robots and the user where both alternatively adjust to each other to maximize the system’s utility. To understand human movement, we recorded motion sequence of several tasks for different subjects using motion capture system. Through analysis of human motion data, we extracted a general model by rule-based approach. On the other hand, since such tasks are not feasible with static models, we investigate the potential benefit of supervised online learning in the task of online action classi?cation and Deep Learning in the task of acquiring human motion. Finally, developed system was extended to show its potential effect in ergonomics and in assist robotics.     

Multiple targets enclosure by robotic swarm

Target enclosure by autonomous robots is useful for many practical applications, for example, surveillance of disaster sites. Scalability is important for autonomous robots because a larger group is more robust against breakdown, accidents, and failure. However, since the traditional models have discussed only the cases in which minimum number of robots enclose a single target, there has been no study on the utilization of the redundant number of robots. In this paper, to achieve a highly scalable target enclosure model about the number of target to enclose, we introduce swarm based task assignment capability to Takayama’s enclosure model. The original model discussed only single target environment but it is well suited for applying to the environments with multiple targets. We show the robots can enclose the targets without predefined position assignment by analytic discussion based on switched systems and a series of computer simulations. As a consequence of this property, the proposed robots can change their target according to the criterion about robot density while they enclose multiple targets.     

Distributed cohesive configuration controller for a swarm with low-cost platforms

This study presents a cohesive configuration controller for distributed space coverage by a swarm of robots. The goal is to build a dense, convex network that is robust against disconnection while robots are flocking with only incomplete knowledge about the network. The controller is an integrated framework of two different algorithms. First, we present a boundary force algorithm: physics-based swarm intelligence that borrows the concept of surface tension force between liquid molecules. The combination of such a force with conventional flocking produces a convex and dense configuration without knowledge of the complete geometry of a robot network. Second, robots distributively determine when a configuration is on the verge of disconnection by identifying a local articulation point—a region where the removal of a single robot will change the local topology. When such a point is detected, robots switch their behavior to clustering, which aggregates them around the vulnerable region to remove every articulation point and retain a connected configuration. Finally, we introduced an index that objectively represents the level of risk of a robot configuration against the massive fragmentation, called vulnerability index. We provide theoretical performance analyses of each algorithm and validate the results with simulations and experiments using a set of low-cost robots.     

Modeling and stability analysis of social foraging swarms in multi-obstacle environment

In this article we specify an individual-based foraging swarm (i.e., group of agents) model with individuals that move in an n-dimensional multi-obstacle environment. The motion of each individual (i) is determined by three factors: i) attraction to the local object position (xˉio+ ) which is decided by the local information about the individuals’ position that individual i can find; ii) repulsion from the other individuals on short distances; and iii) attraction to the global object position (xgoal ) or repulsion from the obstacles in the environment. The emergent behavior of the swarm motion is the result of a balance between inter-individual interaction and the simultaneous interactions of the swarm members with their environment. We study the stability properties of the collective behavior of the swarm based on Lyapunov stability theory. The simulations show that the swarm can converge to goal regions and diverge from obstacle regions of the environment while maintaining cohesive.     

Survey on frontiers of language and robotics

ABSTRACT

The understanding and acquisition of a language in a real-world environment is an important task for future robotics services. Natural language processing and cognitive robotics have both been focusing on the problem for decades using machine learning. However, many problems remain unsolved despite significant progress in machine learning (such as deep learning and probabilistic generative models) during the past decade. The remaining problems have not been systematically surveyed and organized, as most of them are highly interdisciplinary challenges for language and robotics. This study conducts a survey on the frontier of the intersection of the research fields of language and robotics, ranging from logic probabilistic programming to designing a competition to evaluate language understanding systems. We focus on cognitive developmental robots that can learn a language from interaction with their environment and unsupervised learning methods that enable robots to learn a language without hand-crafted training data.     

Get in touch: cooperative decision making based on robot-to-robot collisions

We demonstrate the ability of a swarm of autonomous micro-robots to perform collective decision making in a dynamic environment. This decision making is an emergent property of decentralized self-organization, which results from executing a very simple bio-inspired algorithm. This algorithm allows the robotic swarm to choose from several distinct light sources in the environment and to aggregate in the area with the highest illuminance. Interestingly, these decisions are formed by the collective, although no information is exchanged by the robots. The only communicative act is the detection of robot-to-robot encounters. We studied the performance of the robotic swarm under four environmental conditions and investigated the dynamics of the aggregation behaviour as well as the flexibility and the robustness of the solutions. In summary, we can report that the tested robotic swarm showed two main characteristic features of swarm systems: it behaved flexible and the achieved solutions were very robust. This was achieved with limited individual sensor abilities and with low computational effort on each single robot in the swarm.     

Towards temporal verification of swarm robotic systems

A robot swarm is a collection of simple robots designed to work together to carry out some task. Such swarms rely on the simplicity of the individual robots; the fault tolerance inherent in having a large population of identical robots; and the self-organised behaviour of the swarm as a whole. Although robot swarms present an attractive solution to demanding real-world applications, designing individual control algorithms that can guarantee the required global behaviour is a difficult problem. In this paper we assess and apply the use of formal verification techniques for analysing the emergent behaviours of robotic swarms. These techniques, based on the automated analysis of systems using temporal logics, allow us to analyse whether all possible behaviours within the robot swarm conform to some required specification. In particular, we apply model-checking, an automated and exhaustive algorithmic technique, to check whether temporal properties are satisfied on all the possible behaviours of the system. We target a particular swarm control algorithm that has been tested in real robotic swarms, and show how automated temporal analysis can help to refine and analyse such an algorithm.     

Coordinating microscopic robots in viscous fluids

Multiagent control provides strategies for aggregating microscopic robots (“nanorobots”) in fluid environments relevant for medical applications. Unlike larger robots, viscous forces and Brownian motion dominate the behavior. Examples range from modified microorganisms (programmable bacteria) to future robots using ongoing developments in molecular computation, sensors and motors. We evaluate controls for locating a cell-sized area emitting a chemical into a moving fluid with parameters corresponding to chemicals released in response to injury or infection in small blood vessels. These control methods are passive Brownian motion, following the chemical concentration gradient, and cooperative behaviors in which some robots use acoustic signals to guide others to the chemical source. Control performance is evaluated using diffusion equations to describe the robot motions and control state transitions. The quantitative results show these control techniques are feasible approaches to the task with trade-offs among fabrication difficulty, response speed, false positive detection rate and energy use. Controlled aggregation at chemically distinctive locations could be useful for sensitive diagnosis, selective changes to biological tissues and forming structures using previous proposals for multiagent control of modular robots.     

Building Terrain-Covering Ant Robots: A Feasibility Study

Robotics researchers have studied robots that can follow trails laid by other robots. We, on the other hand, study robots that leave trails in the terrain to cover closed terrain repeatedly. How to design such ant robots has so far been studied only theoretically for gross robot simplifications. In this article, we describe for the first time how to build physical ant robots that cover terrain and test their design both in realistic simulation environments and on a Pebbles III robot. We show that the coverage behavior of our ant robots can be modeled with a modified version of node counting, a real-time search method. We then report on first experiments that we performed to understand their efficiency and robustness in situations where some ant robots fail, they are moved without realizing this, the trails are of uneven quality, and some trails are destroyed. Finally, we report the results of a large-scale simulation experiment where ten ant robots covered a factory floor of 25 by 25 meters repeatedly over 85 hours without getting stuck.     

基于随机行走的群机器人二维地图构建

机器人在未知环境自主探索时,需要快速准确地获取环境地图信息。针对高效探索和未知环境的地图构建问题,将随机行走算法应用于群机器人的探索中,机器人模拟布朗运动,对搜索区域建图。然后,改进了布朗运动算法,通过设置机器人随机行走时的最大旋转角度,来避免机器人重复性地搜索一个区域,使机器人在相同时间内探索更多的区域,提高机器人的搜索效率。最后,通过搭载激光雷达的多个移动机器人进行了仿真实验,实验分析了最大转角增量、机器人数量以及机器人运动步数对搜索区域的影响。     

Tracking natural trails with swarm‐based visual saliency

This paper proposes a model for trail detection and tracking that builds upon the observation that trails are salient structures in the robot's visual field. Due to the complexity of natural environments, the straightforward application of bottom‐up visual saliency models is not sufficiently robust to predict the location of trails. As for other detection tasks, robustness can be increased by modulating the saliency computation based on a priori knowledge about which pixel‐wise visual features are most representative of the object being sought. This paper proposes the use of the object's overall layout as the primary cue instead, as it is more stable and predictable in natural trails. Bearing in mind computational parsimony and detection robustness, this knowledge is specified in terms of perception‐action rules, which control the behavior of simple agents performing as a swarm to compute the saliency map of the input image. For the purpose of tracking, multiframe evidence about the trail location is obtained with a motion‐compensated dynamic neural field. In addition, to reduce ambiguity between the trail and trail‐like distractors, a simple appearance model is learned online and used to influence the agents' activity. Experimental results on a large data set reveal the ability of the model to produce a success rate on the order of 97% at 20 Hz. The model is shown to be robust in situations where previous models would fail, such as when the trail does not emerge from the lower part of the image or when it is considerably interrupted. © 2012 Wiley Periodicals, Inc.     

Symmetry and symmetry-breaking in multiagent behavior

This paper presents some results about properties of the spatial behavior of systems consisting of many artificial agents (robots). The general goal is to understand how the complexity of group behavior is related to individual behavior, and how differences arise and can be grounded. We argue that symmetry and symmetry-breaking are essential factors for the organization of spatial group behavior. They can serve as guiding principles for the construction of agents for real-world environments. This work was presented, in part, at the Second International Symposium on Artifical Life and Robotics, Oita, Japan, February 18–20, 1997