Evolution of low-complexity neural controllers based on multiobjective evolution

In this paper, we present a new method based on multiobjective evolutionary algorithms to evolve low-complexity neural controllers for agents that have to perform multiple tasks simultaneously. In our method, each task and the structure of the neural controller are considered as separated objective functions. We compare the results of two different encoding schemes: (1) connectionist encoding, and (2) node-based encoding. The results show that multiobjective evolution can be successfully applied to generate low-complexity neural controllers. In addition, node-based encoding outperformed connectionist encoding in terms of agent performance and the robustness of the neural controller. This work was presented in part at the 12th International Symposium on Artificial Life and Robotics, Oita, Japan, January 25–27, 2007     

Humanoid robot arm performance optimization using multi objective evolutionary algorithm

As humanoid robots are expected to operate in human environments they are expected to perform a wide range of tasks. Therefore, the robot arm motion must be generated based on the specific task. In this paper we propose an optimal arm motion generation satisfying multiple criteria. In our method, we evolved neural controllers that generate the humanoid robot arm motion satisfying three different criteria; minimum time, minimum distance and minimum acceleration. The robot hand is required to move from the initial to the final goal position. In order to compare the performance, single objective GA is also considered as an optimization tool. Selected neural controllers from the Pareto solution are implemented and their performance is evaluated. Experimental investigation shows that the evolved neural controllers performed well in the real hardware of the mobile humanoid robot platform.     

Co-manipulation with a library of virtual guiding fixtures

Virtual guiding fixtures constrain the movements of a robot to task-relevant trajectories, and have been successfully applied to, for instance, surgical and manufacturing tasks. Whereas previous work has considered guiding fixtures for single tasks, in this paper we propose a library of guiding fixtures for multiple tasks, and propose methods for (1) creating and adding guides based on machine learning; (2) selecting guides on-line based on probabilistic implementation of guiding fixtures; (3) refining existing guides based on an incremental learning method. We demonstrate in an industrial task that a library of guiding fixtures provides an intuitive haptic interface for joint human–robot completion of tasks, and improves performance in terms of task execution time, mental workload and errors.     

A neural network approach for a robot task sequencing problem

This paper presents a neural network approach with successful implementation for the robot task-sequencing problem. The problem addresses the sequencing of tasks comprising loading and unloading of parts into and from the machines by a material-handling robot. The performance criterion is to minimize a weighted objective of the total robot travel time for a set of tasks and the tardiness of the tasks being sequenced. A three-phased parallel implementation of the neural network algorithm on Thinking Machine's CM-5 parallel computer is also presented which resulted in a dramatic increase in the speed of finding solutions. To evaluate the performance of the neural network approach, a branch-and-bound method and a heuristic procedure have been developed for the problem. The neural network method is shown to give good results and is especially useful for solving large problems on a parallel-computing platform.     

Evolution of adaptive synapses: robots with fast adaptive behavior in new environments.

This paper is concerned with adaptation capabilities of evolved neural controllers. We propose to evolve mechanisms for parameter self-organization instead of evolving the parameters themselves. The method consists of encoding a set of local adaptation rules that synapses follow while the robot freely moves in the environment. In the experiments presented here, the performance of the robot is measured in environments that are different in significant ways from those used during evolution. The results show that evolutionary adaptive controllers solve the task much faster and better than evolutionary standard fixed-weight controllers, that the method scales up well to large architectures, and that evolutionary adaptive controllers can adapt to environmental changes that involve new sensory characteristics (including transfer from simulation to reality and across different robotic platforms) and new spatial relationships.     

基于多约束条件的机器人抓取策略学习方法

针对机器人在多类别物体不同任务下的抓取决策问题,提出基于多约束条件的抓取策略学习方法.该方法以抓取对象特征和抓取任务属性为机器人抓取策略约束,通过映射人类抓取习惯规划抓取模式,并采用物体方向包围盒(OBB)建立机器人抓取规则,建立多约束条件的抓取模型.利用深度径向基(DRBF)网络模型结合减聚类算法(SCM)实现抓取策略的学习,两种算法的结合旨在提高学习鲁棒性与精确性.搭建以Refiex 1型灵巧手和AUBO六自由度机械臂组成的实验平台,对多类别物体进行抓取实验.实验结果表明,所提出方法使机器人有效学习到对多物体不同任务的最优抓取策略,具有良好的抓取决策能力.     

An algorithm for cooperative task allocation in scalable,constrained multiple robot systems

The current trends in the robotics field have led to the development of large-scale multiple robot systems, and they are deployed for complex missions. The robots in the system can communicate and interact with each other for resource sharing and task processing. Many of such systems fail despite the availability of necessary resources. The major reason for this is their poor coordination mechanism. Task planning, which involves task decomposition and task allocation, is paramount in the design of coordination and cooperation strategies of multiple robot systems. Task allocation mechanism allocates the task in a mission to the robots by maximizing the overall expected performance, and thereby reducing the total allocation cost for the team. In this paper, we formulate a heuristic search-based task allocation algorithm for the task processing in heterogeneous multiple robot system, by maximizing the efficiency in terms of both communication and processing cost. We assume a set of decomposed tasks of a mission, which needs to be allocated to the robots. The near-optimal allocation schemes are found using the proposed peer structure algorithm for the given problem, where the number of the tasks is more than the robots present in the system. The cost function is the summation of static overhead cost of robots, assignment cost, and the communication cost between the dependent tasks, if they are assigned to different robots. Experiments are performed to verify the effectiveness of the algorithm by comparing it with the existing methods in terms of computational time and quality of solution. The experimental results show that the proposed algorithm performs the best under different problem scales. This proves that the algorithm can be scaled for larger system and it can work for dynamic multiple robot system.     

Recognition and path planning strategy for autonomous navigation in the elevator environment

This paper presents a robust and reliable method for a mobile robot to get on/off an elevator in a multistory building. Getting on/off the elevator requires the robot to perform two different tasks: a recognition task and a navigation task. First, we propose a recognition algorithm for the elevator buttons and status so that the robot reacts flexibly to the current elevator status. We first apply an adaptive threshold to the current image in order to get a binary image. Then we extract the candidates of the buttons and the floor number after preliminary filtering. Ambiguous candidates are rejected using an artificial neural network, and a matching method is applied to finally recognize the call buttons, destination floor buttons, moving direction and current location of the elevator. Second, we suggest a path planning algorithm to navigate into and out of the elevator without any collision. By constructing an occupancy grid map and computing a target function, we find the best position for the robot to get on the elevator. Then we plan an optimal path to the best position using a potential field method. Experiments were carried out in several simulated and real environments including empty, crowd and blocked scenarios. The approach presented here has been found to allow the robot to navigate in the elevator without collisions.     

Two-stage adaptive robot position/force control using fuzzy reasoning and neural networks

Many studies have been performed on the position/force control of robot manipulators. Since the desired position and force required to realize certain tasks are usually designated in the operational space, the controller should adapt itself to an environment and generate the control force vector in the operational space. On the other hand, the friction of each joint of a robot manipulator is a serious problem since it impedes control accuracy. Therefore, the friction should be effectively compensated for in order to realize precise control of robot manipulators. Recently, soft computing techniques (fuzzy reasoning, neural networks and genetic algorithms) have been playing an important role in the control of robots. Applying the fuzzy-neuro approach (a combination of fuzzy reasoning and neural networks), learning/adaptation ability and human knowledge can be incorporated into a robot controller. In this paper, we propose a two-stage adaptive robot manipulator position/force control method in which the uncertain/unknown dynamic of the environment is compensated for in the task space and the joint friction is effectively compensated for in the joint space using soft computing techniques. The effectiveness of the proposed control method was evaluated by experiments.     

A feature selection method for a sample-based stochastic policy

Stochastic policy gradient methods have been applied to a variety of robot control tasks such as robot’s acquisition of motor skills because they have an advantage in learning in high-dimensional and continuous feature spaces by combining some heuristics like motor primitives. However, when we apply one of them to a real-world task, it is difficult to represent the task well by designing the policy function and the feature space due to the lack of enough prior knowledge about the task. In this research, we propose a method to extract a preferred feature space autonomously to achieve a task using a stochastic policy gradient method for a sample-based policy. We apply our method to a control of linear dynamical system and the computer simulation result shows that a desirable controller is obtained and that the performance of the controller is improved by the feature selection.     

Realization of highly energy efficient pick-and-place tasks using resonance-based robot motion control

In resonance-based robot motion control, the potential energy stored in mechanical elasticity is effectively utilized to generate robot motion; therefore, high energy efficiency is expected when the desired motions are periodic. However, several practical problems arise when resonance-based robot motion control is applied to pick-and-place tasks. This paper proposes methods to solve these problems. First, experiments are conducted to verify the energy efficiency of resonance-based robot motion control when a SCARA-type robot performs a pick-and-place task. Second, as a previous method can only treat two points and increasing of the number pick-and-place points is critically important to conduct actual pick-and-place tasks in factories, methods are proposed to increase the number of the pick and/or place points. Finally, the performance of the proposed method is experimentally investigated.     

云计算平台中非定期任务并行调度仿真

由于云计算平台的动态不确定性和非定期任务调度本身的复杂性,使得非定期任务调度过程中的耗时长和负载不均等问题很难得到有效解决.针对上述问题,提出一种非定期任务并行调度方法,并应用到云计算中.通过多方面考虑云平台客户非定期任务的截止时间底线、调度估算等并行调度约束条件和各种可用资源的性能参数,对非定期任务调度的多目标约束条...     

基于场景理解与改进型BUG算法的移动机器人避障

针对现有移动机器人在视觉避障上存在的局限,将深度学习算法和路径规划技术相结合,提出了一种基于深层卷积神经网络和改进Bug算法的机器人避障方法;该方法采用多任务深度卷积神经网络提取道路图像特征,实现图像分类和语义分割任务;其次,基于语义分割结果构建栅格地图,并将图像分类结果与改进的Bug算法相结合,搜索出最优避障路径;同时,为降低冗余计算,设计了特征对比结构来对避免对重复计算的特征信息,保障机器人在实际应用中实时性;通过实验结果表明,所提方法有效的平衡了多视觉任务的精度与效率,并能准确规划出安全的避障路径,辅助机器人完成导航避障。     

A method to learn the inverse kinematics of multi-link robots by evolving neuro-controllers

A general method to learn the inverse kinematic of multi-link robots by means of neuro-controllers is presented. We can find analytical solutions for the most used and well-known robots in the literature. However, these solutions are specific to a particular robot configuration and are not generally applicable to other robot morphologies. The proposed method is general in the sense that it is independent of the robot morphology. The method is based on the evolutionary computation paradigm and works obtaining incrementally better neuro-controllers. Furthermore, the proposed method solves some specific issues in robotic neuro-controller learning: it avoids any neural network learning algorithm which relies on the classical supervised input-target learning scheme and hence it lets to obtain neuro-controllers without providing targets. It can converge beyond local optimal solutions, which is one of the main drawbacks of some neural network training algorithms based on gradient descent when applied to highly redundant robot morphologies. Furthermore, using learning algorithms such as the neuro-evolution of augmenting topologies it is also possible to learn the neural network topology which is a common source of empirical testing in neuro-controllers design. Finally, experimental results are provided when applying the method to two multi-link robot learning tasks and a comparison between structural and parametric evolutionary strategies on neuro-controllers is shown.     

Hybrid moment/position control of a parallel robot

In this paper, a hybrid moment/position controller in task space is proposed for tasks involving a contact between a robot and its environment. We consider a contour-tracking task performed by a six DOF (Degrees Of Freedom) parallel robot. The task space dynamic model of the robot in contact with its environment, seen as a black box, is estimated by a MLP-NN (MultiLayer Perceptron Neural Network). The neural network non-linearity is treated using Taylor series expansion. An adaptation algorithm of the neural parameters resulting from a closed-loop stability analysis is proposed. The performance of the proposed controller is validated on the C5 parallel robot by considering two different environments: rigid and compliant.     

Cognitive Architecture for Robust Adaptive Control of Robots in a Team

The objective of this paper is to present a cognitive architecture thatutilizes three different methodologies for adaptive, robust control ofrobots behaving intelligently in a team. The robots interact within a worldof objects, and obstacles, performing tasks robustly, while improving theirperformance through learning. The adaptive control of the robots has beenachieved by a novel control system. The Tropism-based cognitive architecturefor the individual behavior of robots in a colony is demonstrated throughexperimental investigation of the robot colony. This architecture is basedon representation of the likes and dislikes of the robots. It is shown thatthe novel architecture is not only robust, but also provides the robots withintelligent adaptive behavior. This objective is achieved by utilization ofthree different techniques of neural networks, machine learning, and geneticalgorithms. Each of these methodologies are applied to the tropismarchitecture, resulting in improvements in the task performance of the robotteam, demonstrating the adaptability and robustness of the proposed controlsystem.     

Incremental reinforcement learning for multi-objective robotic tasks

Recently reinforcement learning has been widely applied to robotic tasks. However, most of these tasks hide more than one objective. In these cases, the construction of a reward function is a key and difficult issue. A typical solution is combining the multiple objectives into one single-objective reward function. However, quite often this formulation is far from being intuitive, and the learning process might converge to a behaviour far from what we need. Another alternative to face these multi-objective tasks is to use what is called transfer learning. In this case, the idea is to reuse the experience gained after the learning of an objective to learn a new one. Nevertheless, the transfer affects only to the learned policy, leaving out other gained information that might be relevant. In this paper, we propose a different approach to learn problems with more than one objective. In particular, we describe a two-stage approach. During the first stage, our algorithm will learn a policy compatible with a main goal at the same time that it gathers relevant information for a subsequent search process. Once this is done, a second stage will start, which consists of a cyclical process of small perturbations and stabilizations, and which tries to avoid degrading the performance of the system while it searches for a new valid policy but that also optimizes a sub-objective. We have applied our proposal for the learning of the biped walking. We have tested it on a humanoid robot, both on simulation and on a real robot.     

基于边缘节点的深度神经网络任务分配方法

随着边缘计算的发展,边缘节点的计算规模不断增加,现有的边缘设备难以搭载深度神经网络模型,网络通信与云端服务器承受着巨大压力。为解决上述问题,通过对Roofline模型进行改进,借助新模型对边缘设备的性能与网络环境进行动态评估。根据评估指标,对神经网络模型进行分离式拆分,部分计算任务分配给边缘节点完成,云端服务器结合节点返回数据完成其它任务。该方法基于节点自身性能与网络环境,进行动态任务分配,具有一定兼容性与鲁棒性。实验结果表明,基于边缘节点的深度神经网络任务分配方法可在不同环境中利用设备的闲置性能,大幅度降低中心服务器的计算负载。     

Towards selective attention: generating image features by learning a visuo-motor map

Robots require a form of visual attention to perform a wide range of tasks effectively. Existing approaches specify in advance the image features and attention control scheme required for a given robot to perform a specific task. However, to cope with different tasks in a dynamic environment, a robot should be able to construct its own attentional mechanisms. This paper presents a method that a robot can use to generating image features by learning a visuo-motor map. The robot constructs the visuo-motor map from training data, and the map constrains both the generation of image features and the estimation of state vectors. The resulting image features and state vectors are highly task-oriented. The learned mechanism is attentional in the sense that it determines what information to select from the image to perform a task. We examine robot experiments using the proposed method for indoor navigation and scoring soccer goals.