Integrating information from multiple sensors enhances the performance of autonomous vehicle perception systems. However, current multimodal 3D object detection methods focus on unifying modalities into a bird’s-eye view (BEV) representation, which overlooks the inherent characteristics of camera perspective view (PV), where 2D detection performance significantly surpasses that of state-of-the-art 3D detectors. In this paper, we propose V-Fusion, a high-quality 2D detection-enhanced multimodal BEV object detection method. By leveraging the 2D priors of PV, we construct 3D query proposals that complement BEV 3D queries. To address the modal discrepancy in generating 3D queries from 2D priors, we propose a depth-robust 2D-to-3D query generation strategy. Additionally, we introduce a novel geometry-constrained self-attention mechanism to enhance the interaction of BEV 3D queries and employ an additional set of learnable 3D queries to account for potentially missed objects. Notably, V-Fusion achieves 74.1 NDS performance on the challenging nuScenes dataset, outperforming SparseFusion in 1.0 NDS and offering comparable inference speed.

In recent years, the combination of Reinforcement Learning (RL) and Evolutionary Algorithm (EA) has been widely studied. Evolutionary Reinforcement Learning (ERL) and its various variants have also been proposed. However, most ERL methods focus on evolving the parameters of the policy network which are high-dimensional. The low sample efficiency of EA imposes limitations when optimizing high-dimensional spaces, thus significantly restricting the performance of such methods. Additionally, most existing methods only focus on the quality of actions, without considering the novelty of actions, which can lead to insufficient exploration. To address these two problems, we propose Quality-Diversity Driven Action Swarm Evolution (ASE-QD). Building upon the off-policy RL method TD3, ASEQD adopts EA to evolve the action space which is low-dimensional and introduces a diversity metric for actions. This metric is used together with the quality metric in EA to select the best actions. We test ASE-QD on various control locomotion tasks and evaluate the model’s robustness by introducing delayed rewards to the environment. The results demonstrate that ASE-QD outperforms TD3 and several state-of-the-art ERL methods in terms of average return, convergence speed, and robustness. Furthermore, we conduct ablation experiments, which confirm that the introduction of diversity metrics indeed enhances the model’s adaptability to different environments.

Leveraging learned strategies in unfamiliar scenarios is fundamental to human intelligence. In reinforcement learning, rationally reusing the policies acquired from other tasks or human experts is critical for tackling problems that are difficult to learn from scratch. In this work, we present a framework called Selective Myopic bEhavior Control (SMEC), which results from the insight that the short-term behaviors of prior policies are sharable across tasks. By evaluating the behaviors of prior policies via a hybrid value function architecture, SMEC adaptively aggregates the sharable short-term behaviors of prior policies and the long-term behaviors of the task policy, leading to coordinated decisions. Empirical results on a collection of manipulation and locomotion tasks demonstrate that SMEC outperforms existing methods, and validate the ability of SMEC to leverage related prior policies.

With the rapid development of intelligent electric vehicles (IEVs), an increasing number of algorithms are being deployed on software platforms, progressively leading to the formation of automotive operating systems (OS). This paper introduces firstly the status of automotive OS and analyzes their development trends. As an emerging technology, automotive OS is a safety-critical system that plays a vital role in driving safety. To address the challenges associated with ensuring the safety of automotive OS, this paper proposes systematically safety measures across multiple perspectives, including software architecture, time protection, memory protection, software monitoring, and communication. Furthermore, a system safety mode is proposed from an engineering implementation perspective. Finally, experimental results validate the feasibility of deploying multiple communication protocols within automotive OS, effectively addressing the challenges posed by big data and high-concurrency demands in IEVs.

Task allocation and path planning in multi-robot systems have been widely applied in various storage and logistics scenarios. In recent years, an increasing number of learning-based methods have been employed in the field of path planning, enabling robots to plan paths with limited perception. However, current research integrating multi-robot task allocation with path planning still relies on methods that require centralized planning, and the cost estimation methods used for task allocation do not consider the time consumed for mutual collision avoidance among multiple robots, limiting both the scale of task handling and the accuracy of task allocation. To address this challenge, this paper proposes a framework that utilizes reinforcement learning (RL) and genetic algorithms (GA) to integrate task allocation and path planning. By implementing this framework, tasks can be allocated more appropriately and collision-free optimal paths can be generated. Experimental results demonstrate that the proposed algorithm achieves higher efficiency in complex multi-robot scenarios by effectively managing congestion and optimizing task allocation.

Abstract Graph products have been extensively applied to model complex networks with striking properties observed in real-world complex systems. In this paper, we study the hitting times for random walks on a class of graphs generated iteratively by edge corona product. We first derive recursive solutions to the eigenvalues and eigenvectors of the normalized adjacency matrix associated with the graphs. Based on these results, we further obtain interesting quantities about hitting times of random walks, providing iterative formulas for two-node hitting time, as well as closed-form expressions for the Kemeny’s constant defined as a weighted average of hitting times over all node pairs, as well as the arithmetic mean of hitting times of all pairs of nodes.

Jointly optimising both the body and brain of a robot is known to be a challenging task, especially when attempting to evolve designs in simulation that will subsequently be built in the real world. To address this, it is increasingly common to combine evolution with a learning algorithm that can either improve the inherited controllers of new offspring to fine tune them to the new body design or learn them from scratch. In this paper an approach is proposed in which a robot is specified indirectly by two compositional pattern producing networks (CPPN) encoded in a single genome, one which encodes the brain and the other the body. The body part of the genome is evolved using an evolutionary algorithm (EA), with an individual learning algorithm (also an EA) applied to the inherited controller to improve it. The goal of this paper is to determine how to utilise the results of learning process most effectively to improve task performance of the robot. Specifically, three variants are investigated: (1) evolution of the body+controller only; (2) a learning algorithm is applied to the inherited controller with the learned fitness assigned to the genome; (3) learning is applied and the genome is updated with the learned controller, as well as being assigned the learned fitness. Experiments are performed in three different scenarios chosen to favour different bodies and locomotion patterns. It is shown that better performance can be obtained using learning but only if the learned controller is inherited by the offspring.

The formation of opinions is fundamentally a network-based process, where the opinions of individuals in a social network exchange, evolve, and eventually convergence towards a specific distribution. However, this dynamic process may be susceptible to manipulation by adversarial entities, who aim to maliciously influence the opinion formulation. The adversary may engage in extensive influence campaigns, disseminating misinformation among populations, thereby potentially destabilizing societies. It is thus of significance to develop strategies to defend against such attacks, which are essential for fostering a healthy environment for information sharing, social deliberation, and opinion formation. In this paper, we investigate a scenario wherein an external adversary aims to maliciously alter the opinions of a general social graph. This is achieved by targeting several selected nodes, referred to as followers. Concurrently, we explore a counter-strategy, aiming to negate the influence of the adversary with malicious intents. This involves identifying a subset of nodes to act as followers of a defending leader, thereby minimizing the adversary’s impact. Since this problem can be framed as a non-increasing supermodular minimization problem, we develop a (1-1/e) approximation greedy algorithm consequently. Moreover, to overcome the computation challenge for large-scale networks, we establish an efficient approximation to the key quantity of the greedy algorithm. This refinement significantly enhances computational efficiency and scalability, making the algorithm applicable to networks with millions of nodes. Extensive simulation results on various real-world networks demonstrate the superior performance of our improved algorithm over existing algorithms and other baseline schemes based on centrality measures. In particular, our improved algorithm scales to networks of considerable size, with negligible sacrifice on the quality of solutions.

A signed graph offers richer information than an unsigned graph, since it describes both collaborative and competitive relationships in social networks. In this paper, we study opinion dynamics on a signed graph, based on the Friedkin-Johnsen model. We first interpret the equilibrium opinion in terms of a defined random walk on an augmented signed graph, by representing the equilibrium opinion of every node as a combination of all nodes’ internal opinions, with the coefficient of the internal opinion for each node being the difference of two absorbing probabilities. We then quantify some relevant social phenomena and express them in terms of the$\ell _{2}$norms of vectors. We also design a nearly-linear time signed Laplacian solver for assessing these quantities, by establishing a connection between the absorbing probability of random walks on a signed graph and that on an associated unsigned graph. We further study the opinion optimization problem by changing the initial opinions of a fixed number of nodes, which can be optimally solved in cubic time. We provide a nearly-linear time algorithm with an error guarantee to approximately solve the problem. Finally, we execute extensive experiments on sixteen real-life signed networks, which show that both of our algorithms are effective and efficient, and are scalable to massive graphs with over 20 million nodes.

In this paper, we propose a new ant colony optimization algorithm, called learning-based neural ant colony optimization (LN-ACO), which incorporates an "intelligent ant". This intelligent ant contains a convolutional neural network pre-trained on a large set of instances which is able to predict the selection probabilities of the set of possible choices at each step of the algorithm. The intelligent ant is capable of generating a solution based on knowledge learned during training, but also guides other 'traditional' ants in improving their choices during the search. As the search progresses, the intelligent ant is also influenced by the pheromones accumulated by the colony, leading to better solutions. The key idea is that if tasks or instances share common features either in terms of their search landscape or solutions, then information learned by solving one instance can be applied to substantially accelerate the search on another. We evaluate the proposed algorithm on two public datasets and one real-world test set in the path planning domain. The results demonstrate that LN-ACO is competitive in its search capability compared to other ACO methods, with a significant improvement in convergence speed.

Learning a policy with great generalization to unseen environments remains challenging but critical in visual reinforcement learning. Despite the success of augmentation combination in the supervised learning generalization, naively applying it to visual RL algorithms may damage the training efficiency, suffering from serve performance degradation. In this paper, we first conduct qualitative analysis and illuminate the main causes: (i) high-variance gradient magnitudes and (ii) gradient conflicts existed in various augmentation methods. To alleviate these issues, we propose a general policy gradient optimization framework, named Conflict-aware Gradient Agreement Augmentation (CG2A), and better integrate augmentation combination into visual RL algorithms to address the generalization bias. In particular, CG2A develops a Gradient Agreement Solver to adaptively balance the varying gradient magnitudes, and introduces a Soft Gradient Surgery strategy to alleviate the gradient conflicts. Extensive experiments demonstrate that CG2A significantly improves the generalization performance and sample efficiency of visual RL algorithms.

In this paper, we study the opinion maximization problem for the leader-follower DeGroot model of opinion dynamics in a social network modelled by a directed graph with n nodes, where a small number of nodes are competing leader nodes with binary opposing opinions 0 or 1, and the rest are follower nodes. We address the problem of maximizing the overall opinion by adding k ⇐ n new edges, where each edge is incident to a 1-leader and a follower. We prove that the objective function is monotone and submodular, and then propose a deterministic greedy algorithm with an approximation ratio (1-1 over e) and O(n3) running time. We then develop a fast sampling algorithm based on l-truncated absorbing random walks and sample-materialization techniques, which has sublinear time complexity O(kn1/2 log3/2 n/ε3) for any error parameter ε > 0. We provide extensive experiments on real networks to evaluate the performance of our algorithms. The results show that for undirected graphs our fast sampling algorithm outperforms the state-of-the-art method in terms of efficiency and effectiveness. While for directed graphs our fast sampling algorithm is as effective as our deterministic greedy algorithm, both of which are much better than the baseline strategies. Moreover, our fast algorithm is scalable to large directed graphs with over 41 million nodes.

Searching solutions for the task with sparse or deceptive rewards is a fundamental problem in Evolutionary Algorithms (EA) and Reinforcement Learning (RL). Existing methods in RL have been proposed to enhance the exploration by encouraging agents to obtain novel states. However, solely seeking a single local optimal solution could be insufficient for the tasks with the deceptive local optima. Novelty-Search (NS) and Quality-Diversity (QD) have shown promising results for finding diverse solutions with different behavioral characteristics. However, manually defining the task-specific behavior description limits these methods to low-dimensional tasks. This paper presents Dynamics-aware Novelty Search with Behavior Repulsion (DANSBR), a hybrid algorithm that evolves high-performing solutions by introducing a generalized novelty measurement and a bidirectional gradient-based mutation operator based on the Quality-Diversity paradigm. The novelty of a single solution is defined as the prediction error of an approximate dynamic model in the task-agnostic behavior space. The mutation operator drives the solution to behave differently or obtain better performance in a sample-eficient manner. As a result of better exploration, our approach outperforms several baselines on high-dimensional continuous control tasks with sparse rewards. Empirical results also demonstrate that DANSBR improves the performance on the task with deceptive rewards.

Recently, adversarial attacks have been applied in visual object tracking to deceive deep trackers by injecting imperceptible perturbations into video frames. However, previous work only generates the video-specific perturbations, which restricts its application scenarios. In addition, existing attacks are difficult to implement in reality due to the real-time of tracking and the re-initialization mechanism. To address these issues, we propose an offline universal adversarial attack called Efficient Universal Shuffle Attack. It takes only one perturbation to cause the tracker malfunction on all videos. To improve the computational efficiency and attack performance, we propose a greedy gradient strategy and a triple loss to efficiently capture and attack model-specific feature representations through the gradients. Experimental results show that EUSA can significantly reduce the performance of state-of-the-art trackers on OTB2015 and VOT2018.

For the drone motion planning problem in dense obstacle scenarios, we introduce a trajectory generation method based on imitation learning that does not require the establish-ment of a local map, which greatly increases the planning speed. This method utilizes only onboard sensors and depth camera perception. We specially made the Imitation Learning Planning-Drones (ILP-Drones) dataset for training. The kinodynamic and smoothness of the generated trajectory are improved with local nonlinear optimization. The uniform B-Spline parameterization is adopted to allocate a reasonable time interval for the generated trajectory. Ultimately, our method is able to plan high quality trajectories with excellent collision avoidance ability within mil-liseconds. This is demonstrated by comparative experiments with various advanced algorithms. At the same time, the flexibility and adaptability of our method are demonstrated by ablation experiments with different number of predicted points and different simulation environments.

With the increasing application of versatile robotics, the need for end-users to teach robotic tasks via visual/video demonstrations in different environments is increasing fast. One possible method is meta-learning. However, most meta-learning methods are tailored for image classification or just focus on teaching the robot what to do, resulting in a limited ability of the robot to adapt to the real world. Thus, we propose a novel yet efficient model-agnostic meta-learning framework based on task-contrastive learning to teach the robot what to do and what not to do through positive and negative demonstrations. Our approach divides the learning procedure from visual/video demonstrations into three parts. The first part distinguishes between positive and negative demonstrations via task-contrastive learning. The second part emphasizes what the positive demo is doing, and the last part predicts what the robot needs to do. Finally, we demonstrate the effectiveness of our meta-learning approach on 1) two standard public simulated benchmarks and 2) real-world placing experiments using a UR5 robot arm, significantly outperforming current related state-of-the-art methods.

Robots are becoming ubiquitous: from vacuum cleaners to driverless cars, there is a wide variety of applications, many with potential safety hazards. The work presented in this paper proposes a set of constructs suitable for both modelling robotic applications and supporting verification via model checking and theorem proving. Our goal is to support roboticists in writing models and applying modern verification techniques using a language familiar to them. To that end, we present RoboChart, a domain-specific modelling language based on UML, but with a restricted set of constructs to enable a simplified semantics and automated reasoning. We present the RoboChart metamodel, its well-formedness rules, and its process-algebraic semantics. We discuss verification based on these foundations using an implementation of RoboChart and its semantics as a set of Eclipse plug-ins called RoboTool.

Robot software controllers are often concurrent and time critical, and requires modern engineering approaches for validation and verification. With this motivation, we have developed a tool and techniques for graphical modelling with support for automatic generation of underlying mathematical definitions for model checking. It is possible to check automatically both general properties, like absence of deadlock, and specific application properties. We cater both for timed and untimed modelling and verification. Our approach has been tried in examples used in a variety of robotic applications.

Recently, a new class of machine learning algorithms has emerged, where models and discriminators are generated in a competitive setting. The most prominent example is Generative Adversarial Networks (GANs). In this paper we examine how these algorithms relate to the Turing test, and derive what - from a Turing perspective - can be considered their defining features. Based on these features, we outline directions for generalizing GANs - resulting in the family of algorithms referred to as Turing Learning. One such direction is to allow the discriminators to interact with the processes from which the data samples are obtained, making them "interrogators", as in the Turing test. We validate this idea using two case studies. In the first case study, a computer infers the behavior of an agent while controlling its environment. In the second case study, a robot infers its own sensor configuration while controlling its movements. The results confirm that by allowing discriminators to interrogate, the accuracy of models is improved.

This paper proposes a strategy for transporting a large object to a goal using a large number of mobile robots that are significantly smaller than the object. The robots only push the object at positions where the direct line of sight to the goal is occluded by the object. This strategy is fully decentralized and requires neither explicit communication nor specific manipulation mechanisms. We prove that it can transport any convex object in a planar environment. We implement this strategy on the e-puck robotic platform and present systematic experiments with a group of 20 e-pucks transporting three objects of different shapes. The objects were successfully transported to the goal in 43 out of 45 trials. When using a mobile goal, teleoperated by a human, the object could be navigated through an environment with obstacles. We also tested the strategy in a 3-D environment using physics-based computer simulation. Due to its simplicity, the transport strategy is particularly suited for implementation on microscale robotic systems.

We propose a coevolutionary approach for learning the behavior of animals, or agents, in collective groups. The approach requires a replica that resembles the animal under investigation in terms of appearance and behavioral capabilities. It is able to identify the rules that govern the animals in an autonomous manner. A population of candidate models, to be executed on the replica, compete against a population of classifiers. The replica is mixed into the group of animals and all individuals are observed. The fitness of the classifiers depends solely on their ability to discriminate between the replica and the animals based on their motion over time. Conversely, the fitness of the models depends solely on their ability to 'trick' the classifiers into categorizing them as an animal. Our approach is metric-free in that it autonomously learns how to judge the resemblance of the models to the animals. It is shown in computer simulation that the system successfully learns the collective behaviors of aggregation and of object clustering. A quantitative analysis reveals that the evolved rules approximate those of the animals with a good precision.

This paper proposes a method that allows a machine to infer the behavior of an animal in a fully automatic way. In principle, the machine does not need any prior information about the behavior. It is able to modify the environmental conditions and observe the animal; therefore it can learn about the animal through controlled interaction. Using a competitive coevolutionary approach, the machine concurrently evolves animats, that is, models to approximate the animal, as well as classifiers to discriminate between animal and animat. We present a proof-of-concept study conducted in computer simulation that shows the feasibility of the approach. Moreover, we show that the machine learns significantly better through interaction with the animal than through passive observation. We discuss the merits and limitations of the approach and outline potential future directions.