arXiv subjects

Chengshu Li

Publications and source records attributed to Chengshu Li.

43 records · Page 3Linked to original sources

arXiv2021

BEHAVIOR: Benchmark for Everyday Household Activities in Virtual, Interactive, and Ecological Environments

We introduce BEHAVIOR, a benchmark for embodied AI with 100 activities in simulation, spanning a range of everyday household chores such as cleaning, maintenance, and food preparation. These activities are designed to be realistic, diverse, and complex, aiming to reproduce the challenges that agents must face in the real world. Building such a benchmark poses three fundamental difficulties for each activity: definition (it can differ by time, place, or person), instantiation in a simulator, and evaluation. BEHAVIOR addresses these with three innovations. First, we propose an object-centric, predicate logic-based description language for expressing an activity's initial and goal conditions, enabling generation of diverse instances for any activity. Second, we identify the simulator-agnostic features required by an underlying environment to support BEHAVIOR, and demonstrate its realization in one such simulator. Third, we introduce a set of metrics to measure task progress and efficiency, absolute and relative to human demonstrators. We include 500 human demonstrations in virtual reality (VR) to serve as the human ground truth. Our experiments demonstrate that even state of the art embodied AI solutions struggle with the level of realism, diversity, and complexity imposed by the activities in our benchmark. We make BEHAVIOR publicly available at behavior.stanford.edu to facilitate and calibrate the development of new embodied AI solutions.

cs.RO↗

arXiv2021

ReLMoGen: Leveraging Motion Generation in Reinforcement Learning for Mobile Manipulation

Many Reinforcement Learning (RL) approaches use joint control signals (positions, velocities, torques) as action space for continuous control tasks. We propose to lift the action space to a higher level in the form of subgoals for a motion generator (a combination of motion planner and trajectory executor). We argue that, by lifting the action space and by leveraging sampling-based motion planners, we can efficiently use RL to solve complex, long-horizon tasks that could not be solved with existing RL methods in the original action space. We propose ReLMoGen -- a framework that combines a learned policy to predict subgoals and a motion generator to plan and execute the motion needed to reach these subgoals. To validate our method, we apply ReLMoGen to two types of tasks: 1) Interactive Navigation tasks, navigation problems where interactions with the environment are required to reach the destination, and 2) Mobile Manipulation tasks, manipulation tasks that require moving the robot base. These problems are challenging because they are usually long-horizon, hard to explore during training, and comprise alternating phases of navigation and interaction. Our method is benchmarked on a diverse set of seven robotics tasks in photo-realistic simulation environments. In all settings, ReLMoGen outperforms state-of-the-art Reinforcement Learning and Hierarchical Reinforcement Learning baselines. ReLMoGen also shows outstanding transferability between different motion generators at test time, indicating a great potential to transfer to real robots.

cs.AI↗

arXiv2020

Coupled wire construction of a topological phase with chiral tricritical Ising edge modes

Tricritical Ising (TCI) phase transition is known to occur in several interacting spin and Majorana fermion models and is described in terms of a supersymmetric conformal field theory (CFT) with central charge $c=7/10$. The field content of this CFT is highly nontrivial and includes among its primary fields the Fibonacci anyon, making it of potential interest to strategies seeking to implement fault-tolerant topological quantum computation with non-Abelian phases of matter. In this paper we explore the possibility that a TCI CFT can occur at the edge of a gapped two-dimensional topological state as a stable phase. We discuss a possible realization of this 2D phase based on a coupled-wire construction using the Grover-Sheng-Vishwanath chain model of Majorana zero modes coupled to Ising spins which is known to undergo the TCI phase transition. From the combined analysis using mean-field theory, conformal field theory and density matrix renormalization group (DMRG) on 2- and 4-leg ladders, we find that the left- and right-moving gapless TCI modes become spatially separated and reside on two opposite edges of the system, forming a precursor of the required 2D topological phase.

cond-mat.str-el↗

arXiv2019

Supersymmetry in an interacting Majorana model on the kagome lattice

We construct a supersymmetric model of interacting Majorana fermions on the kagome lattice. In the infinite-coupling limit, the model exhibits an extensively degenerate ground state manifold separated in two topological sectors, in addition to two parity supersectors. An exact solution for thin-torus geometries allows us to analytically construct the entire zero-energy ground state manifold. Upon inclusion of hopping terms with energy scale $g$, the supersymmetry is spontaneously broken with the ground state energy density scaling as $g^4$. We also briefly discuss the non-interacting limit of the model, which exhibits a zero-energy flat band and Majorana Chern bands.

cond-mat.str-el↗

arXiv2019

HRL4IN: Hierarchical Reinforcement Learning for Interactive Navigation with Mobile Manipulators

Most common navigation tasks in human environments require auxiliary arm interactions, e.g. opening doors, pressing buttons and pushing obstacles away. This type of navigation tasks, which we call Interactive Navigation, requires the use of mobile manipulators: mobile bases with manipulation capabilities. Interactive Navigation tasks are usually long-horizon and composed of heterogeneous phases of pure navigation, pure manipulation, and their combination. Using the wrong part of the embodiment is inefficient and hinders progress. We propose HRL4IN, a novel Hierarchical RL architecture for Interactive Navigation tasks. HRL4IN exploits the exploration benefits of HRL over flat RL for long-horizon tasks thanks to temporally extended commitments towards subgoals. Different from other HRL solutions, HRL4IN handles the heterogeneous nature of the Interactive Navigation task by creating subgoals in different spaces in different phases of the task. Moreover, HRL4IN selects different parts of the embodiment to use for each phase, improving energy efficiency. We evaluate HRL4IN against flat PPO and HAC, a state-of-the-art HRL algorithm, on Interactive Navigation in two environments - a 2D grid-world environment and a 3D environment with physics simulation. We show that HRL4IN significantly outperforms its baselines in terms of task performance and energy efficiency. More information is available at https://sites.google.com/view/hrl4in.

cs.LG↗

arXiv2018

Majorana-Hubbard model on the honeycomb lattice

Phase diagram of a Hubbard model for Majorana fermions on the honeycomb lattice is explored using a combination of field theory, renormalization group and mean-field arguments, as well as exact numerical diagonalization. Unlike the previously studied versions of the model we find that even weak interactions break symmetries and lead to interesting topological phases. We establish two topologically nontrivial phases at weak coupling, one gapped with chiral edge modes, and the other gapless with anti-chiral edge modes. At strong coupling a mapping onto a novel frustrated spin-${1\over 2}$ model suggests a highly entangled spin liquid ground state.

cond-mat.str-el↗

arXiv2018

Family of Sachdev-Ye-Kitaev models motivated by experimental considerations

Several condensed-matter platforms have been proposed recently to realize the Sachdev-Ye-Kitaev (SYK) model in their low-energy limit. In these proposed realizations, the characteristic SYK behavior is expected to occur under certain assumptions about the underlying physical system that (i) render all bilinear terms small compared to four-fermion interactions and (ii) ensure that the coupling constants are approximately all-to-all and independent random variables. In this work we explore, both analytically and numerically, the family of models that arises when we relax these assumptions in ways motivated by real physical systems. By relaxing (i) and allowing large bilinear terms, we obtain a novel, exactly-solvable cousin of the SYK model. It exhibits two distinct phases separated by a quantum phase transition characterized by a power-law, $\sim |ω|^{-1/3}$ scaling of the low-energy spectral density, despite being a non-interacting model. By relaxing (ii), we obtain close relatives of the SYK model which exhibit interesting behaviors, including a chaotic non-Fermi liquid phase with continuously varying fermion scaling dimension, and a phase transition to a disordered Fermi liquid as a function of interaction range and disorder length scale.

cond-mat.str-el↗