We propose a new task-specification language for Markov decision processes that is designed to be an improvement over reward functions by being environment independent. The variant of Linear Temporal Logic (LTL) extended probabilistic specifications in way permits approximations learned finite time. provide several small environments demonstrate the advantages our geometric LTL (GLTL) and illustrate how it can used specify standard reinforcement-learning tasks straightforwardly.

We consider synthesis of control policies that maximize the probability satisfying given temporal logic specifications in unknown, stochastic environments. model interaction between system and its environment as a Markov decision process (MDP) with initially unknown transition probabilities. The solution we develop builds on so-called model-based probably approximately correct (PAC-MDP) methodology. algorithm attains an $\varepsilon$-approximately optimal policy $1-\delta$ using samples...

Soft pneumatic actuators enable robots to interact safely with complex environments, but often suffer from imprecise control and unpredictable dynamics. This letter addresses these challenges through the use of model reference adaptive control, which modulates input plant ensure that it behaves similarly a dynamic model. We standardize performance soft eliminate their nonlinear behavior. implement an controller chosen for its simplicity efficiency, study ability this force different behave...

It is often desirable to capture and map semantic information of an environment during simultaneous localization mapping (SLAM). Such can enable a robot better distinguish places with similar low-level geometric visual features perform high-level tasks that use about objects be manipulated environments navigated. While SLAM has gained increasing attention, there little research on semantic-level data association based objects, i.e., object-level association. In this paper, we propose novel...

Snakes are a remarkable source of inspiration for mobile search-and-rescue robots. Their unique slender body structure and multiple modes locomotion well-suited to movement in narrow passages other difficult terrain. The design, manufacturing, modeling, control techniques soft robotics make it possible imitate the structure, mechanical properties, gaits snakes, opening up new possibilities research. Building on our track record contributions this area, article presents robotic snake made...

Large language models (LLMs) have enabled remarkable advances in automated task-solving with multi-agent systems. However, most existing LLM-based approaches rely on predefined agents to handle simple tasks, limiting the adaptability of collaboration different scenarios. Therefore, we introduce AutoAgents, an innovative framework that adaptively generates and coordinates multiple specialized build AI team according tasks. Specifically, AutoAgents couples relationship between tasks roles by...

Ensuring that AI systems make strategic decisions aligned with the specified preferences in adversarial sequential interactions is a critical challenge for developing trustworthy systems, especially when environment stochastic and players' incomplete leave some outcomes unranked. We study problem of synthesizing preference-satisfying strategies two-player games on graphs where players have opposite (possibly incomplete) over set temporal goals. represent these goals using linear logic finite...

Loop closure is necessary for correcting errors accumulated in simultaneous localization and mapping (SLAM) unknown environments. However, conventional loop methods based on low-level geometric or image features may cause high ambiguity by not distinguishing similar scenarios. Thus, incorrect closures can occur. Though semantic 2D information considered some literature to detect closures, there little work that compares 3D scenes as an integral part of a SLAM system. This letter introduces...

We propose a synthesis method of switching control policies for class shared autonomy systems in which authority is held by either human operator or an autonomous controller based on the state overall system. The objective to optimize system performance measured probability satisfying specification linear temporal logic, while ensuring reasonable workload operator. builds upon construction abstract model given from set components modeled Markov decision processes, are capable capturing...

In this paper, we present a new locomotion control method for soft robot snakes. Inspired by biological snakes, our architecture is composed of two key modules: A reinforcement learning (RL) module achieving adaptive goal-tracking behaviors with changing goals, and central pattern generator (CPG) system Matsuoka oscillators generating stable diverse patterns. The modules are interconnected into closed-loop system: RL module, analogizing the region located in midbrain vertebrate animals,...

This paper describes localization and navigation in an indoor environment with transparent objects. Although LIDAR proved to be a popular choice, its inability detect glass surface requires another sensor ensure capability will work all kind of environment. In this paper, sonar is used complement laser due hard surfaces low cost. The experiment shows that better map can achieved when these two sensors are used.

With the increasing sophistication of attacks on cyber-physical systems, deception has emerged as an effective tool to improve system security and safety by obfuscating attacker's perception. In this letter, we present a solution deceptive game in which control agent is satisfy Boolean objective specified co-safe temporal logic formula presence adversary. The intentionally introduces asymmetric information create payoff misperception, manifests misperception labeling function model. Thus,...

Geometric fracture assembly presents a challenging practical task in archaeology and 3D computer vision. Previous methods have focused solely on assembling fragments based semantic information, which has limited the quantity of objects that can be effectively assembled. Therefore, there is need to develop scalable framework for geometric without relying information. To improve effectiveness fractures we propose co-creation space comprising several assemblers capable gradually unambiguously...

Temporal logic provides a rigorous and expressive way to model specifications for cyber-physical systems (CPS). Such systems, which are playing increasingly important roles in various domains, also subject malicious attacks initiated by intelligent adversaries. Malicious may prevent the system from satisfying desired specification. In this article, we introduce control synthesis method minimize deviations specification due We consider scenario where adversary has limited capability of...

We study planning problems where autonomous agents operate inside environments that are subject to uncertainties and not fully observable. Partially observable Markov decision processes (POMDPs) a natural formal model capture such problems. Because of the potentially huge or even infinite belief space in POMDPs, synthesis with safety guarantees is, general, computationally intractable. propose an approach aims circumvent this difficulty: scenarios can be partially simulated virtual...

Urban public transit planning is crucial in reducing traffic congestion and enabling green transportation. However, there no systematic way to integrate passengers' personal preferences routes schedules so as achieve high occupancy rates efficiency gain of ride-sharing. In this paper, we take the first step tp exact from history data. We propose a data-driven method construct Markov decision process model that characterizes passengers making sequential choices, bus routes, subway lines,...

This paper studies the deployment of joint moving target defense (MTD) and deception against multi-stage cyber-attacks. Given system equipped with MTD that randomizes between different configurations, we investigate how to allocate a bounded number sensors in each configuration optimize probability detecting attack before attacker achieves its objective. Specifically, two types are considered: intrusion detectors observable by stealthy not attacker. We propose two-step optimization-based...

We introduce TACO, an open-source, large-scale code generation dataset, with a focus on the optics of algorithms, designed to provide more challenging training dataset and evaluation benchmark in field models. TACO includes competition-level programming questions that are challenging, enhance or evaluate problem understanding reasoning abilities real-world scenarios. There 25433 1000 coding problems test set, as well up 1.55 million diverse solution answers. Moreover, each several...

In this work we present a framework that is capable of accurately representing soft robotic actuators in multiphysics environment real-time. We propose constraint-based dynamics model 1-dimensional pneumatic actuator accounts for internal pressure forces, as well the effect latency and damping under inflation deflation demonstrate its accuracy full snake with composition multiple 1D actuators. verify our model's static deformation dynamic locomotion open-loop control experiments. To achieve...

The fast pace of global urbanization is drastically changing the population distributions over world, which leads to significant changes in geographical densities. Such turn alter underlying power demand time, and drive substations become over-supplied (demand << capacity) or under-supplied ≈ capacity). In this paper, we make first attempt investigate problem substation-user assignment by analyzing large-scale grid data. We develop a Scalable Power User Assignment (SPUA) framework, that...