arXiv subjects

Katia Sycara

Publications and source records attributed to Katia Sycara.

At least 73 records · Page 4Linked to original sources

arXiv2023

Reconfigurable Robot Control Using Flexible Coupling Mechanisms

Reconfigurable robot swarms are capable of connecting with each other to form complex structures. Current mechanical or magnetic connection mechanisms can be complicated to manufacture, consume high power, have a limited load-bearing capacity, or can only form rigid structures. In this paper, we present our low-cost soft anchor design that enables flexible coupling and decoupling between robots. Our asymmetric anchor requires minimal force to be pushed into the opening of another robot while having a strong pulling force so that the connection between robots can be secured. To maintain this flexible coupling mechanism as an assembled structure, we present our Model Predictive Control (MPC) frameworks with polygon constraints to model the geometric relationship between robots. We conducted experiments on the soft anchor to obtain its force profile, which informed the three-bar linkage model of the anchor in the simulations. We show that the proposed mechanism and MPC frameworks enable the robots to couple, decouple, and perform various behaviors in both the simulation environment and hardware platform. Our code is available at https://github.com/ZoomLabCMU/puzzlebot_anchor . Video is available at https://www.youtube.com/watch?v=R3gFplorCJg .

cs.RO↗

arXiv2023

Introspective Action Advising for Interpretable Transfer Learning

Transfer learning can be applied in deep reinforcement learning to accelerate the training of a policy in a target task by transferring knowledge from a policy learned in a related source task. This is commonly achieved by copying pretrained weights from the source policy to the target policy prior to training, under the constraint that they use the same model architecture. However, not only does this require a robust representation learned over a wide distribution of states -- often failing to transfer between specialist models trained over single tasks -- but it is largely uninterpretable and provides little indication of what knowledge is transferred. In this work, we propose an alternative approach to transfer learning between tasks based on action advising, in which a teacher trained in a source task actively guides a student's exploration in a target task. Through introspection, the teacher is capable of identifying when advice is beneficial to the student and should be given, and when it is not. Our approach allows knowledge transfer between policies agnostic of the underlying representations, and we empirically show that this leads to improved convergence rates in Gridworld and Atari environments while providing insight into what knowledge is transferred.

cs.LG↗

arXiv2023

Explainable Action Advising for Multi-Agent Reinforcement Learning

Action advising is a knowledge transfer technique for reinforcement learning based on the teacher-student paradigm. An expert teacher provides advice to a student during training in order to improve the student's sample efficiency and policy performance. Such advice is commonly given in the form of state-action pairs. However, it makes it difficult for the student to reason with and apply to novel states. We introduce Explainable Action Advising, in which the teacher provides action advice as well as associated explanations indicating why the action was chosen. This allows the student to self-reflect on what it has learned, enabling advice generalization and leading to improved sample efficiency and learning performance - even in environments where the teacher is sub-optimal. We empirically show that our framework is effective in both single-agent and multi-agent scenarios, yielding improved policy returns and convergence rates when compared to state-of-the-art methods

cs.AI↗

arXiv2023

Sample-Efficient Learning of Novel Visual Concepts

Despite the advances made in visual object recognition, state-of-the-art deep learning models struggle to effectively recognize novel objects in a few-shot setting where only a limited number of examples are provided. Unlike humans who excel at such tasks, these models often fail to leverage known relationships between entities in order to draw conclusions about such objects. In this work, we show that incorporating a symbolic knowledge graph into a state-of-the-art recognition model enables a new approach for effective few-shot classification. In our proposed neuro-symbolic architecture and training methodology, the knowledge graph is augmented with additional relationships extracted from a small set of examples, improving its ability to recognize novel objects by considering the presence of interconnected entities. Unlike existing few-shot classifiers, we show that this enables our model to incorporate not only objects but also abstract concepts and affordances. The existence of the knowledge graph also makes this approach amenable to interpretability through analysis of the relationships contained within it. We empirically show that our approach outperforms current state-of-the-art few-shot multi-label classification methods on the COCO dataset and evaluate the addition of abstract concepts and affordances on the Visual Genome dataset.

cs.CV↗

arXiv2023

Distributed Multirobot Control for Non-Cooperative Herding

In this paper, we consider the problem of protecting a high-value area from being breached by sheep agents by crafting motions for dog robots. We use control barrier functions to pose constraints on the dogs' velocities that induce repulsion in the sheep relative to the high-value area. This paper extends the results developed in our prior work on the same topic in three ways. Firstly, we implement and validate our previously developed centralized herding algorithm on many robots. We show herding of up to five sheep agents using three dog robots. Secondly, as an extension to the centralized approach, we develop two distributed herding algorithms, one favoring feasibility while the other favoring optimality. In the first algorithm, we allocate a unique sheep to a unique dog, making that dog responsible for herding its allocated sheep away from the protected zone. We provide feasibility proof for this approach, along with numerical simulations. In the second algorithm, we develop an iterative distributed reformulation of the centralized algorithm, which inherits the optimality (i.e. budget efficiency) from the centralized approach. Lastly, we conduct real-world experiments of these distributed algorithms and demonstrate herding of up to five sheep agents using five dog robots.

cs.RO↗

arXiv2023

Graph-based Simultaneous Coverage and Exploration Planning for Fast Multi-robot Search

In large unknown environments, search operations can be much more time-efficient with the use of multi-robot fleets by parallelizing efforts. This means robots must efficiently perform collaborative mapping (exploration) while simultaneously searching an area for victims (coverage). Previous simultaneous mapping and planning techniques treat these problems as separate and do not take advantage of the possibility for a unified approach. We propose a novel exploration-coverage planner which bridges the mapping and search domains by growing sets of random trees rooted upon a pose graph produced through mapping to generate points of interest, or tasks. Furthermore, it is important for the robots to first prioritize high information tasks to locate the greatest number of victims in minimum time by balancing coverage and exploration, which current methods do not address. Towards this goal, we also present a new multi-robot task allocator that formulates a notion of a hierarchical information heuristic for time-critical collaborative search. Our results show that our algorithm produces 20% more coverage efficiency, defined as average covered area per second, compared to the existing state-of-the-art. Our algorithms and the rest of our multi-robot search stack is based in ROS and made open source

cs.RO↗

arXiv2023

On the Role of Emergent Communication for Social Learning in Multi-Agent Reinforcement Learning

Explicit communication among humans is key to coordinating and learning. Social learning, which uses cues from experts, can greatly benefit from the usage of explicit communication to align heterogeneous policies, reduce sample complexity, and solve partially observable tasks. Emergent communication, a type of explicit communication, studies the creation of an artificial language to encode a high task-utility message directly from data. However, in most cases, emergent communication sends insufficiently compressed messages with little or null information, which also may not be understandable to a third-party listener. This paper proposes an unsupervised method based on the information bottleneck to capture both referential complexity and task-specific utility to adequately explore sparse social communication scenarios in multi-agent reinforcement learning (MARL). We show that our model is able to i) develop a natural-language-inspired lexicon of messages that is independently composed of a set of emergent concepts, which span the observations and intents with minimal bits, ii) develop communication to align the action policies of heterogeneous agents with dissimilar feature models, and iii) learn a communication policy from watching an expert's action policy, which we term `social shadowing'.

cs.LG↗

arXiv2023

Concept Learning for Interpretable Multi-Agent Reinforcement Learning

Multi-agent robotic systems are increasingly operating in real-world environments in close proximity to humans, yet are largely controlled by policy models with inscrutable deep neural network representations. We introduce a method for incorporating interpretable concepts from a domain expert into models trained through multi-agent reinforcement learning, by requiring the model to first predict such concepts then utilize them for decision making. This allows an expert to both reason about the resulting concept policy models in terms of these high-level concepts at run-time, as well as intervene and correct mispredictions to improve performance. We show that this yields improved interpretability and training stability, with benefits to policy performance and sample efficiency in a simulated and real-world cooperative-competitive multi-agent game.

cs.LG↗

arXiv2023

Predicting Out-of-Distribution Error with Confidence Optimal Transport

Out-of-distribution (OOD) data poses serious challenges in deployed machine learning models as even subtle changes could incur significant performance drops. Being able to estimate a model's performance on test data is important in practice as it indicates when to trust to model's decisions. We present a simple yet effective method to predict a model's performance on an unknown distribution without any addition annotation. Our approach is rooted in the Optimal Transport theory, viewing test samples' output softmax scores from deep neural networks as empirical samples from an unknown distribution. We show that our method, Confidence Optimal Transport (COT), provides robust estimates of a model's performance on a target domain. Despite its simplicity, our method achieves state-of-the-art results on three benchmark datasets and outperforms existing methods by a large margin.

cs.LG↗

arXiv2023

Interpretable Learned Emergent Communication for Human-Agent Teams

Learning interpretable communication is essential for multi-agent and human-agent teams (HATs). In multi-agent reinforcement learning for partially-observable environments, agents may convey information to others via learned communication, allowing the team to complete its task. Inspired by human languages, recent works study discrete (using only a finite set of tokens) and sparse (communicating only at some time-steps) communication. However, the utility of such communication in human-agent team experiments has not yet been investigated. In this work, we analyze the efficacy of sparse-discrete methods for producing emergent communication that enables high agent-only and human-agent team performance. We develop agent-only teams that communicate sparsely via our scheme of Enforcers that sufficiently constrain communication to any budget. Our results show no loss or minimal loss of performance in benchmark environments and tasks. In human-agent teams tested in benchmark environments, where agents have been modeled using the Enforcers, we find that a prototype-based method produces meaningful discrete tokens that enable human partners to learn agent communication faster and better than a one-hot baseline. Additional HAT experiments show that an appropriate sparsity level lowers the cognitive load of humans when communicating with teams of agents and leads to superior team performance.

cs.LG↗

arXiv2022

ARC - Actor Residual Critic for Adversarial Imitation Learning

Adversarial Imitation Learning (AIL) is a class of popular state-of-the-art Imitation Learning algorithms commonly used in robotics. In AIL, an artificial adversary's misclassification is used as a reward signal that is optimized by any standard Reinforcement Learning (RL) algorithm. Unlike most RL settings, the reward in AIL is $differentiable$ but current model-free RL algorithms do not make use of this property to train a policy. The reward is AIL is also shaped since it comes from an adversary. We leverage the differentiability property of the shaped AIL reward function and formulate a class of Actor Residual Critic (ARC) RL algorithms. ARC algorithms draw a parallel to the standard Actor-Critic (AC) algorithms in RL literature and uses a residual critic, $C$ function (instead of the standard $Q$ function) to approximate only the discounted future return (excluding the immediate reward). ARC algorithms have similar convergence properties as the standard AC algorithms with the additional advantage that the gradient through the immediate reward is exact. For the discrete (tabular) case with finite states, actions, and known dynamics, we prove that policy iteration with $C$ function converges to an optimal policy. In the continuous case with function approximation and unknown dynamics, we experimentally show that ARC aided AIL outperforms standard AIL in simulated continuous-control and real robotic manipulation tasks. ARC algorithms are simple to implement and can be incorporated into any existing AIL implementation with an AC algorithm. Video and link to code are available at: https://sites.google.com/view/actor-residual-critic.

cs.LG↗

arXiv2022

Towards True Lossless Sparse Communication in Multi-Agent Systems

Communication enables agents to cooperate to achieve their goals. Learning when to communicate, i.e., sparse (in time) communication, and whom to message is particularly important when bandwidth is limited. Recent work in learning sparse individualized communication, however, suffers from high variance during training, where decreasing communication comes at the cost of decreased reward, particularly in cooperative tasks. We use the information bottleneck to reframe sparsity as a representation learning problem, which we show naturally enables lossless sparse communication at lower budgets than prior art. In this paper, we propose a method for true lossless sparsity in communication via Information Maximizing Gated Sparse Multi-Agent Communication (IMGS-MAC). Our model uses two individualized regularization objectives, an information maximization autoencoder and sparse communication loss, to create informative and sparse communication. We evaluate the learned communication `language' through direct causal analysis of messages in non-sparse runs to determine the range of lossless sparse budgets, which allow zero-shot sparsity, and the range of sparse budgets that will inquire a reward loss, which is minimized by our learned gating function with few-shot sparsity. To demonstrate the efficacy of our results, we experiment in cooperative multi-agent tasks where communication is essential for success. We evaluate our model with both continuous and discrete messages. We focus our analysis on a variety of ablations to show the effect of message representations, including their properties, and lossless performance of our model.

cs.LG↗

arXiv2022

Control Barrier Functions-based Semi-Definite Programs (CBF-SDPs): Robust Safe Control For Dynamic Systems with Relative Degree Two Safety Indices

In this draft article, we consider the problem of achieving safe control of a dynamic system for which the safety index or (control barrier function (loosely)) has relative degree equal to two. We consider parameter affine nonlinear dynamic systems and assume that the parametric uncertainty is uniform and known a-priori or being updated online through an estimator/parameter adaptation law. Under this uncertainty, the usual CBF-QP safe control approach takes the form of a robust optimization problem. Both the right hand side and left hand side of the inequality constraints depend on the unknown parameter. With the given representation of uncertainty, the CBF-QP safe control ends up being a convex semi-infinite problem. Using two different philosophies, one based on weak duality and another based on the Lossless s-procedure, we arrive at identical SDP formulations of this robust CBF-QP problem. Thus we show that the problem of computing safe controls with known parametric uncertainty can be posed as a tractable convex problem and be solved online. (This is work in progress).

math.OC↗

arXiv2022

The Before, During, and After of Multi-Robot Deadlock

Collision avoidance for multirobot systems is a well-studied problem. Recently, control barrier functions (CBFs) have been proposed for synthesizing controllers that guarantee collision avoidance and goal stabilization for multiple robots. However, it has been noted that reactive control synthesis methods (such as CBFs) are prone to \textit{deadlock}, an equilibrium of system dynamics that causes the robots to stall before reaching their goals. In this paper, we analyze the closed-loop dynamics of robots using CBFs, to characterize controller parameters, initial conditions, and goal locations that invariably lead the system to deadlock. Using tools from duality theory, we derive geometric properties of robot configurations of an $N$ robot system once it is in deadlock and we justify them using the mechanics interpretation of KKT conditions. Our key deductions are that 1) system deadlock is characterized by a force-equilibrium on robots and 2) deadlock occurs to ensure safety when safety is on the brink of being violated. These deductions allow us to interpret deadlock as a subset of the state space, and we show that this set is non-empty and located on the boundary of the safe set. By exploiting these properties, we analyze the number of admissible robot configurations in deadlock and develop a provably-correct decentralized algorithm for deadlock resolution to safely deliver the robots to their goals. This algorithm is validated in simulations as well as experimentally on Khepera-IV robots.

cs.RO↗

arXiv2022

Noncooperative Herding With Control Barrier Functions: Theory and Experiments

In this paper, we consider the problem of protecting a high-value unit from inadvertent attack by a group of agents using defending robots. Specifically, we develop a control strategy for the defending agents that we call "dog robots" to prevent a flock of "sheep agents" from breaching a protected zone. We take recourse to control barrier functions to pose this problem and exploit the interaction dynamics between the sheep and dogs to find dogs' velocities that result in the sheep getting repelled from the zone. We solve a QP reactively that incorporates the defending constraints to compute the desired velocities for all dogs. Owing to this, our proposed framework is composable \textit{i.e.} it allows for simultaneous inclusion of multiple protected zones in the constraints on dog robots' velocities. We provide a theoretical proof of feasibility of our strategy for the one dog/one sheep case. Additionally, we provide empirical results of two dogs defending the protected zone from upto ten sheep averaged over a hundred simulations and report high success rates. We also demonstrate this algorithm experimentally on non-holonomic robots. Videos of these results are available at https://tinyurl.com/4dj2kjwx.

cs.RO↗

arXiv2022

Configuration Control for Physical Coupling of Heterogeneous Robot Swarms

In this paper, we present a heterogeneous robot swarm system that can physically couple with each other to form functional structures and dynamically decouple to perform individual tasks. The connection between robots can be formed with a passive coupling mechanism, ensuring minimum energy consumption during coupling and decoupling behavior. The heterogeneity of the system enables the robots to perform structural enhancement configurations based on specific environmental requirements. We propose a connection-pair oriented configuration control algorithm to form different assemblies. We show experiments of up to nine robots performing the coupling, gap-crossing, and decoupling behaviors.

cs.RO↗

arXiv2022

PuzzleBots: Physical Coupling of Robot Swarms

Robot swarms have been shown to improve the ability of individual robots by inter-robot collaboration. In this paper, we present the PuzzleBots - a low-cost robotic swarm system where robots can physically couple with each other to form functional structures with minimum energy consumption while maintaining individual mobility to navigate within the environment. Each robot has knobs and holes along the sides of its body so that the robots can couple by inserting the knobs into the holes. We present the characterization of knob design and the result of gap-crossing behavior with up to nine robots. We show with hardware experiments that the robots are able to couple with each other to cross gaps and decouple to perform individual tasks. We anticipate the PuzzleBots will be useful in unstructured environments as individuals and coupled systems in real-world applications.

cs.RO↗

arXiv2022

Probe-Based Interventions for Modifying Agent Behavior

Neural nets are powerful function approximators, but the behavior of a given neural net, once trained, cannot be easily modified. We wish, however, for people to be able to influence neural agents' actions despite the agents never training with humans, which we formalize as a human-assisted decision-making problem. Inspired by prior art initially developed for model explainability, we develop a method for updating representations in pre-trained neural nets according to externally-specified properties. In experiments, we show how our method may be used to improve human-agent team performance for a variety of neural networks from image classifiers to agents in multi-agent reinforcement learning settings.

cs.LG↗