As a prevalent field of study in machine learning, intelligent agents can perceive surroundings and make informed decisions. In many research areas such as autopilot systems, undersea explorations, and distributed robotics, researchers have traditionally employed unidirectional systems, which usually rely on perceptions from sensors to controllers, or end-to-end models, which generate actions directly from raw data. Nonetheless, in unidirectional systems, controller efficiency is intrinsically linked to sensor accuracy, which makes unidirectional systems lack a self-improving capability. Meanwhile, compared with functionally separated frameworks, end-to-end methods may have their own limitations in scalability, generality, interoperability, training costs, and so on. To fulfill this gap, we propose a new Coordinated Optimization Framework for Intelligent Agents (COIA). We introduce an inverted optimization channel from controllers to sensors in traditional functionally separated frameworks through communication between devices, enabling closed-loop online evolutive learning. To the best of our knowledge, this paper first presents a universal coordinated optimization framework among supervised learning and RL models, without human labels or intervention. Our method allows heterogeneous agents to autonomously adapt to some special situations in open environments, which forms a basis of networked Artificial General Intelligence (AGI). We design an experimental paradigm of COIA with concrete cases, which shows a significantly large performance margin over unidirectional and end-to-end models. The performance margin grows with task complexity.

As a potential application of Vehicle-to-Everything (V2X) communication, multi-agent collaborative perception has achieved significant success in 3D object detection. While these methods have demonstrated impressive results on standard benchmarks, the robustness of such approaches in the face of complex real-world environments requires additional verification. To bridge this gap, we introduce the first comprehensive benchmark designed to evaluate the robustness of collaborative perception methods in the presence of natural corruptions typical of real-world environments. Furthermore, we propose DSRC, a robustness-enhanced collaborative perception method aiming to learn Density-insensitive and Semantic-aware collaborative Representation against Corruptions. DSRC consists of two key designs: i) a semantic-guided sparse-to-dense distillation framework, which constructs multi-view dense objects painted by ground truth bounding boxes to effectively learn density-insensitive and semantic-aware collaborative representation; ii) a feature-to-point cloud reconstruction approach to better fuse critical collaborative representation across agents. To thoroughly evaluate DSRC, we conduct extensive experiments on real-world and simulated datasets. The results demonstrate that our method outperforms state-of-the-art collaborative perception methods in both clean and corrupted conditions.

Multi-agent collaborative perception enables a single agent to perceive the comprehensive scene by exchanging sensory information using vehicle-to-everything communication. However, the communication resources of real-world communication systems are often limited. It fails to satisfy the real-time transmission of extensive data, which restricts the deployment of collaborative perception. To address the issue, we propose ComSH, a Communication-efficient multi-agent collaborative perception method based on Spatio-temporal Heterogeneity to achieve a trade-off between performance and bandwidth. Specifically, we consider the spatio-temporal heterogeneity to perform feature filtering, thus reducing the redundancy of transmitted features. First, it introduces valuable temporal semantics to enhance the current representation of each agent. Secondly, we consider the spatial confidence of each agent at each location and the relative position to obtain high confidence and complementary features. Eventually, a foreground object adaptive activation module is designed to enrich the visual representation of fused features. Therefore, ComSH enables different agents to transmit their critical and complementary perception information, thus reducing bandwidth consumption. We conduct experiments on collaborative detection tasks on the two datasets. Experimental results demonstrate the ComSH’s superiority and effectiveness.

In artificial intelligence, enhancing agents’ perception and decision-making in physical environments is a key research focus. Perception and control, as essential components of task execution, interact closely and jointly determine system performance. Traditionally, agents use perception models to extract environmental information and control algorithms to generate actions. However, in real-world scenarios, the choice and combination of perception and control models can greatly affect efficiency and accuracy, demanding stronger adaptability and robustness. To study these effects, we design a simulation platform for path planning tasks built on the OpenAI Gym framework. The platform is flexible and scalable, supporting integration and evaluation of diverse perception-control algorithm combinations. Based on this platform, we conduct experiments with different perception-control pairings and further propose a perception optimization mechanism driven by control feedback, namely online evolutive learning mechanism. This mechanism dynamically adjusts the perception model within the closed-loop system, enabling agents to adapt online and respond more effectively to environmental changes. Experimental results show that perception-control combinations yield notable performance differences under identical conditions. Moreover, the proposed optimization mechanism significantly accelerates convergence and improves execution efficiency in non-stationary environments, confirming its effectiveness in enhancing system robustness and adaptability.

As one of the research objects in machine learning, agents are endowed with the ability to perceive and make decisions. In order to meet the needs of different applications, many researchers focus on algorithms for controlling agents. In most current paradigms, the control algorithms take the information of the surrounding environment as input and output actions. However, under the influence of this unidirectional data flow, the upper limit of the performance of control algorithms depends on the accuracy and fit measure of the environment information. Even though recent end-to-end control algorithms take the environmental raw data as input and reduce the influence of perception performance on it, the raw data acquisition method is still fixed, therefore, the performance of these control algorithms is still limited by the environmental information acquisition scheme. Meanwhile, although most control methods have the ability to update parameters online, they usually do not have the ability to optimize the acquisition of environmental information, and there may be unknown situations that are not in the data set used for pre-training. As a result, current methods cannot handle detection inaccuracies and unknown situations, and lack the ability to further improve their own performance online. In order to resolve the deficiency of the traditional control paradigms, we introduce the reverse optimization channel from the controller to the environmental information acquisition scheme to form a new algorithm through communication between devices. Experiments show that our method has significant performance margin and universal online evolutive learning ability, as compared to traditional paradigms.

The ever-changing traffic dynamics make the traditional traffic signal control methods unable to adapt to the environment. Meanwhile, deep reinforcement learning (DRL) has the property of interacting with the environment and adapting to changes in the environment. Therefore, in recent years, researchers have usually solved traffic signal control (TSC) problems through DRL methods. They have not only improved the design of neural networks, but also improved the ability of models to understand traffic conditions and learn corresponding task requests by designing different states and rewards. However, although the existing TSC algorithms based on DRL have proposed many well-designed states and reward strategies, which combinations of states and rewards should be adopted in practice to achieve the performance margin of models remains a question that researchers are seeking the answer to. Therefore, we introduce a general simulation platform to test and compare experimental performance under different combinations of states and rewards. Specifically, we test and analyze the experimental effects under different combinations of multiple traffic states and rewards through various TSC methods with a set of unified model settings. We further design and test some new state representations and reward strategies based on more detailed traffic information. The test results show that when researchers design the state and reward, refining the traffic state like vehicle running condition and making the state and reward match can make the experimental performance better than other combinations in most cases. We hope these results have some implications for the state and reward choice when researchers conduct experiments on TSC problem or other traffic decision management problems.

Vehicular Edge Computing (VEC), as one of the major components of Intelligent Transportation Systems, improves road safety by providing computing services to safety-related applications on vehicles. Currently, the existing fine-grained computing scheduling algorithms are normally designed based on some simple scheduling policies. Due to the heterogeneous nature of tasks offloaded from various applications, they may not effectively satisfy various performance requirements of the real system, thereby leading to the problem that the short-term residual computing power cannot be effectively utilized when computing-costly tasks occupy the server. Therefore, improving the overall system performance and the efficiency of utilizing computing power is a critical issue. Accordingly, in this paper, we study the problem of computing scheduling inside edge servers in VEC, where multiple tasks can be offloaded to Road Side Units (RSUs). We analyze the role played by multiple evaluation metrics in the existing methods for ensuring the quality of service (QoS) and further design a novel online multi-factor aware task offloading algorithm with a hierarchical fine-grained computing scheduling scheme inside the edge server. We evaluate it by conducting intensive simulation tests and comparing the results with some state-of-the-art approaches. Numerical results show that the proposed algorithm outperforms the methods in the control group in different aspects and achieves the best overall performance.

Cellular Automata (CA) models have become the most commonly used tool for simulating urban expansion. To improve the accuracy of CA models, various driving factors like spatial proximity and neighbourhood effects have been explored in previous studies, but the inclusion of these factors does not address the directional differences in urban expansion. To address this issue, this study develops a method to measure urban spatial anisotropy (SA) with respect to 18 variables at both the global and local scales, and integrates all these SA variables into a logistic regression-based CA model. The revised CA model is evaluated with a case study for Huizhou, China. The case study shows that the simulation results for the CA model with SA exhibit 89% overall accuracy; compared to CA models that do not consider SA, the revised CA model can improve precision by 5% on newly developed cells. The consideration of SA in CA models proves promising in improving the accuracy of urban expansion simulations.