In this paper, we present the tidiness score-guided Monte Carlo tree search (TSMCTS), a novel framework designed to address the tabletop tidying up problem using only an RGB-D camera. We address two major problems for tabletop tidying up problem: (1) the lack of public datasets and benchmarks, and (2) the difficulty of specifying the goal configuration of unseen objects. We address the former by presenting the tabletop tidying up (TTU) dataset, a structured dataset collected in simulation. Using this dataset, we train a vision-based discriminator capable of predicting the tidiness score. This discriminator can consistently evaluate the degree of tidiness across unseen configurations, including real-world scenes. Addressing the second problem, we employ Monte Carlo tree search (MCTS) to find tidying trajectories without specifying explicit goals. Instead of providing specific goals, we demonstrate that our MCTS-based planner can find diverse tidied configurations using the tidiness score as a guidance. Consequently, we propose TSMCTS, which integrates a tidiness discriminator with an MCTS-based tidying planner to find optimal tidied arrangements. TSMCTS has successfully demonstrated its capability across various environments, including coffee tables, dining tables, office desks, and bathrooms. The TTU dataset is available at: https://github.com/rllab-snu/TTU-Dataset.

In robotics, the use of Large Language Models (LLMs) is becoming prevalent, especially for understanding human commands. In particular, LLMs are utilized as domain-agnostic task planners for high-level human commands. LLMs are capable of Chain-of-Thought (CoT) reasoning, and this allows LLMs to be task planners. However, we need to consider that modern robots still struggle to perform complex actions, and the domains where robots can be deployed are limited in practice. This leads us to pose a question: If small LMs can be trained to reason in chains within a single domain, would even small LMs be good task planners for the robots? To train smaller LMs to reason in chains, we build `COmmand-STeps datasets' (COST) consisting of high-level commands along with corresponding actionable low-level steps, via LLMs. We release not only our datasets but also the prompt templates used to generate them, to allow anyone to build datasets for their domain. We compare GPT3.5 and GPT4 with the finetuned GPT2 for task domains, in tabletop and kitchen environments, and the result shows that GPT2-medium is comparable to GPT3.5 for task planning in a specific domain. Our dataset, code, and more output samples can be found in https://github.com/Gawon-Choi/small-LMs-Task-Planning

We propose MR.HuBo (Motion Retargeting leveraging a HUman BOdy prior), a cost-effective and convenient method to collect high-quality upper body paired < robot, human > pose data, which is essential for data-driven motion retargeting methods. Unlike existing approaches which collect < robot, human > pose data by converting human MoCap poses into robot poses, our method goes in reverse. We first sample diverse random robot poses, and then convert them into human poses. However, since random robot poses can result in extreme and infeasible human poses, we propose an additional technique to sort out extreme poses by exploiting a human body prior trained from a large amount of human pose data. Our data collection method can be used for any humanoid robots, if one designs or optimizes the system's hyperparameters which include a size scale factor and the joint angle ranges for sampling. In addition to this data collection method, we also present a two-stage motion retargeting neural network that can be trained via supervised learning on a large amount of paired data. Compared to other learning-based methods trained via unsupervised learning, we found that our deep neural network trained with ample high-quality paired data achieved notable performance. Our experiments also show that our data filtering method yields better retargeting results than training the model with raw and noisy data. Our code and video results are available on https://sites.google.com/view/mr-hubo/.

The synthesis of human motion has traditionally been addressed through task-dependent models that focus on specific challenges, such as predicting future motions or filling in intermediate poses conditioned on known key-poses. In this paper, we present a novel task-independent model called UNIMASK-M, which can effectively address these challenges using a unified architecture. Our model obtains comparable or better performance than the state-of-the-art in each field. Inspired by Vision Transformers (ViTs), our UNIMASK-M model decomposes a human pose into body parts to leverage the spatio-temporal relationships existing in human motion. Moreover, we reformulate various pose-conditioned motion synthesis tasks as a reconstruction problem with different masking patterns given as input. By explicitly informing our model about the masked joints, our UNIMASK-M becomes more robust to occlusions. Experimental results show that our model successfully forecasts human motion on the Human3.6M dataset while achieving state-of-the-art results in motion inbetweening on the LaFAN1 dataset for long transition periods.

We hypothesize dance as a motion that forms a visual rhythm from music, where the visual rhythm can be perceived from an optical flow. If an agent can recognize the relationship between visual rhythm and music, it will be able to dance by generating a motion to create a visual rhythm that matches the music. Based on this, we propose a framework for any kind of non-humanoid agents to learn how to dance from human videos. Our framework works in two processes: (1) training a reward model which perceives the relationship between optical flow (visual rhythm) and music from human dance videos, (2) training the non-humanoid dancer based on that reward model, and reinforcement learning. Our reward model consists of two feature encoders for optical flow and music. They are trained based on contrastive learning which makes the higher similarity between concurrent optical flow and music features. With this reward model, the agent learns dancing by getting a higher reward when its action creates an optical flow whose feature has a higher similarity with the given music feature. Experiment results show that generated dance motion can align with the music beat properly, and user study result indicates that our framework is more preferred by humans compared to the baselines. To the best of our knowledge, our work of non-humanoid agents which learn dance from human videos is unprecedented. An example video can be found at https://youtu.be/dOUPvo-O3QY.

It lies in human nature to properly adjust the muscle force to perform a given task successfully. While transferring this control ability to robots has been a big concern among researchers, there is no attempt to make a robot learn how to control the impedance solely based on visual observations. Rather, the research on tele-impedance usually relies on special devices such as EMG sensors, which have less accessibility as well as less generalization ability compared to simple RGB webcams. In this letter, we propose a system for a vision-based tele-impedance control of robots, based on the approximately estimated muscle activation patterns. These patterns are obtained from the proposed deep learning-based model, which uses RGB images from an affordable commercial webcam as inputs. It is remarkable that our model does not require humans to apply any visible markers to their muscles. Experimental results show that our model enables a robot to mimic how humans adjust their muscle force to perform a given task successfully. Although our experiments are focused on tele-impedance control, our system can also provide a baseline for improvement of vision-based learning from demonstration, which would also incorporate the information of variable stiffness control for successful task execution.

The goal of this paper is to estimate object’s 6D pose based on the texture-less dataset. The pose of each projection view is obtained by rendering the 3D model of each object, and then the orientation feature of the object is implicitly represented by the latent space obtained from the RGB image. The 3D rotation of the object is estimated by establishing the codebook based on a template matching architecture. To build the latent space from the RGB images, this paper proposes a network based on a variant Adversarial Autoencoder (Makhzani et al. in Computer Science, 2015). To train the network, we use the dataset without pose annotation, and the encoder and decoder do not have a structural symmetry. The encoder is inspired by the existing model (Yang et al. in proceedings of IJCAI, 2018), (Yang et al. in proceedings 11 of CVPR, 2019) that incorporates the function of feature extraction from two different streams. Based on this network, the latent feature vector that implicitly represents the orientation of the object is obtained from the RGB image. Experimental results show that the method in this paper can realize the 6D pose estimation of the object and the result accuracy is better than the advanced method (Sundermeyer et al. in proceedings of ECCV, 2018).

After many researchers observed fruitfulness from the recent diffusion probabilistic model, its effectiveness in image generation is actively studied these days. In this paper, our objective is to evaluate the potential of diffusion probabilistic models for 3D human motion-related tasks. To this end, this pa-per presents a study of employing diffusion probabilistic models to predict future 3D human motion(s) from the previously observed motion. Based on the Human 3.6M and HumanEva-I datasets, our results show that diffusion probabilistic models are competitive for both single (deterministic) and multiple (stochastic) 3D motion prediction tasks, after finishing a single training process. In addition, we find out that diffusion probabilistic models can offer an attractive compromise, since they can strike the right balance between the likelihood and diversity of the predicted future motions. Our code is publicly available on the project website: https://sites.google.com/view/diffusion-motion-prediction.

Understanding the human-object interactions (HOIs) from a video is essential to fully comprehend a visual scene. This line of research has been addressed by detecting HOIs from images and lately from videos. However, the video-based HOI anticipation task in the third-person view remains understudied. In this paper, we design a framework to detect current HOIs and anticipate future HOIs in videos. We propose to leverage human gaze information since people often fixate on an object before interacting with it. These gaze features together with the scene contexts and the visual appearances of human-object pairs are fused through a spatio-temporal transformer. To evaluate the model in the HOI anticipation task in a multi-person scenario, we propose a set of person-wise multi-label metrics. Our model is trained and validated on the VidHOI dataset, which contains videos capturing daily life and is currently the largest video HOI dataset. Experimental results in the HOI detection task show that our approach improves the baseline by a great margin of 36.3% relatively. Moreover, we conduct an extensive ablation study to demonstrate the effectiveness of our modifications and extensions to the spatio-temporal transformer. Our code is publicly available on https://github.com/nizhf/hoi-prediction-gaze-transformer.

To anticipate how a person would act in the future, it is essential to understand the human intention since it guides the subject towards a certain action. In this paper, we propose a hierarchical architecture which assumes a sequence of human action (low-level) can be driven from the human intention (high-level). Based on this, we deal with long-term action anticipation task in egocentric videos. Our framework first extracts this low- and high-level human information over the observed human actions in a video through a Hierarchical Multi-task Multi-Layer Perceptrons Mixer (H3M). Then, we constrain the uncertainty of the future through an Intention-Conditioned Variational Auto-Encoder (I-CVAE) that generates multiple stable predictions of the next actions that the observed human might perform. By leveraging human intention as high-level information, we claim that our model is able to anticipate more time-consistent actions in the long-term, thus improving the results over the baseline in Ego4D dataset. This work results in the state-of-the-art for Long-Term Anticipation (LTA) task in Ego4D by providing more plausible anticipated sequences, improving the anticipation scores of nouns and actions. Our work ranked first in both CVPR@2022 and ECCV@2022 Ego4D LTA Challenge.

In this paper, we present a proactive robotic voice assistant with a perceive-reason-act loop that carries out pick-and-place operations based on verbal commands. Unlike existing systems, our robot can retrieve a target object not only when the target is explicitly spelled out, but also given an indirect command that implicitly reflects the human intention or emotion. For instance, when the verbal command is “I had a busy day, so I didn’t have much to eat.”, the target object would be something that can help with hunger. To successfully estimate the target object from indirect commands, our framework consists of separate modules for the complete perceive-reason-act loop as follows. First, for perception, it runs an object detector on the robot’s onboard computer to detect all objects in the surroundings and records a verbal command from a microphone. Second, for reasoning, a list of available objects as well as a transcription of the verbal command are integrated into a prompt for a Large Language Model (LLM) in order to identify the target object in the command. Finally, for action, a TurtleBot3 with a 5 DOF robotic arm finds the target object and brings it to the human. Our experiments show that with a properly designed prompt, the robot can identify the correct target object from implicit commands with at most 97% accuracy. In addition, it is shown that the technique of fine-tuning a language model based on the proposed prompt designing process amplifies the performance of the smallest language model by a factor of five. Our data and code are available at https://github.com/bekatan/vafor

Comprehending human motion is a fundamental challenge for developing Human-Robot Collaborative applications. Computer vision researchers have addressed this field by only focusing on reducing error in predictions, but not taking into account the requirements to facilitate its implementation in robots. In this paper, we propose a new model based on Transformer that simultaneously deals with the real time 3D human motion forecasting in the short and long term. Our 2-Channel Transformer (2CH-TR) is able to efficiently exploit the spatio-temporal information of a shortly observed sequence (400ms) and generates a competitive accuracy against the current state-of-the-art. 2CH-TR stands out for the efficient performance of the Transformer, being lighter and faster than its competitors. In addition, our model is tested in conditions where the human motion is severely occluded, demonstrating its robustness in reconstructing and predicting 3D human motion in a highly noisy environment. Our experiment results show that the proposed 2CH-TR outperforms the ST-Transformer, which is another state-of-the-art model based on the Transformer, in terms of reconstruction and prediction under the same conditions of input prefix. Our model reduces in 8.89% the mean squared error of ST-Transformer in short-term prediction, and 2.57% in long-term prediction in Human3.6M dataset with 400ms input prefix.

This paper emphasizes the importance of a robot's ability to refer to its task history, especially when it executes a series of pick-and-place manipulations by following language instructions given one by one. The advantage of referring to the manipulation history can be categorized into two folds: (1) the language instructions omitting details but using expressions referring to the past can be interpreted, and (2) the visual information of objects occluded by previous manipulations can be inferred. For this, we introduce a history-dependent manipulation task which objective is to visually ground a series of language instructions for proper pick-and-place manipulations by referring to the past. We also suggest a relevant dataset and model which can be a baseline, and show that our model trained with the proposed dataset can also be applied to the real world based on the CycleGAN. Our dataset and code are publicly available on the project website: https://sites.google.com/view/history-dependent-manipulation.

In this paper, we propose Hierarchical Action Segmentation Refiner (HASR), which can refine temporal action segmentation results from various models by understanding the overall context of a given video in a hierarchical way. When a backbone model for action segmentation estimates how the given video can be segmented, our model extracts segment-level representations based on frame-level features, and extracts a video-level representation based on the segment-level representations. Based on these hierarchical representations, our model can refer to the overall context of the entire video, and predict how the segment labels that are out of context should be corrected. Our HASR can be plugged into various action segmentation models (MS-TCN, SSTDA, ASRF), and improve the performance of state-of-the-art models based on three challenging datasets (GTEA, 50Salads, and Breakfast). For example, in 50Salads dataset, the segmental edit score improves from 67.9% to 77.4% (MS-TCN), from 75.8% to 77.3% (SSTDA), from 79.3% to 81.0% (ASRF). In addition, our model can refine the segmentation result from the unseen backbone model, which was not referred to when training HASR. This generalization performance would make HASR be an effective tool for boosting up the existing approaches for temporal action segmentation. Our code is available at https://github.com/cotton-ahn/HASR_iccv2021.

In this paper, we present self-supervised shared latent embedding (S3LE), a data-driven motion retargeting method that enables the generation of natural motions in humanoid robots from motion capture data or RGB videos. While it requires paired data consisting of human poses and their corresponding robot configurations, it significantly alleviates the necessity of time-consuming data-collection via novel paired data generating processes. Our self-supervised learning procedure consists of two steps: automatically generating paired data to bootstrap the motion retargeting, and learning a projection-invariant mapping to handle the different expressivity of humans and humanoid robots. Furthermore, our method guarantees that the generated robot pose is collision-free and satisfies position limits by utilizing nonparametric regression in the shared latent space. We demonstrate that our method can generate expressive robotic motions from both the CMU motion capture database and YouTube videos.

This paper proposes a multiple stakeholder perspective model (MSPM) which predicts the future pedestrian trajectory observed from vehicle's point of view. For the vehicle-pedestrian interaction, the estimation of the pedestrian's intention is a key factor. However, even if this interaction is commonly initiated by both the human (pedestrian) and the agent (driver), current research focuses on developing a neural network trained by the data from driver's perspective only. In this paper, we suggest a multiple stakeholder perspective model (MSPM) and apply this model for pedestrian intention prediction. The model combines the driver (stakeholder 1) and pedestrian (stakeholder 2) by separating the information based on the perspective. The dataset from pedestrian's perspective have been collected from the virtual reality experiment, and a network that can reflect perspectives of both pedestrian and driver is proposed. Our model achieves the best performance in the existing pedestrian intention dataset, while reducing the trajectory prediction error by average of 4.48% in the short-term (0.5s) and middle-term (1.0s) prediction, and 11.14% in the long-term prediction (1.5s) compared to the previous state-of-the-art.

This letter proposes a framework which is able to generate a sequence of three-dimensional human dance poses for a given music. The proposed framework consists of three components: a music feature encoder, a pose generator, and a music genre classifier. We focus on integrating these components for generating a realistic 3D human dancing move from music, which can be applied to artificial agents and humanoid robots. The trained dance pose generator, which is a generative autoregressive model, is able to synthesize a dance sequence longer than 1,000 pose frames. Experimental results of generated dance sequences from various songs show how the proposed method generates human-like dancing move to a given music. In addition, a generated 3D dance sequence is applied to a humanoid robot, showing that the proposed framework can make a robot to dance just by listening to music.

In this paper, we propose the Interactive Text2Pickup (IT2P) network for human-robot collaboration which enables an effective interaction with a human user despite the ambiguity in user's commands. We focus on the task where a robot is expected to pick up an object instructed by a human, and to interact with the human when the given instruction is vague. The proposed network understands the command from the human user and estimates the position of the desired object first. To handle the inherent ambiguity in human language commands, a suitable question which can resolve the ambiguity is generated. The user's answer to the question is combined with the initial command and given back to the network, resulting in more accurate estimation. The experiment results show that given unambiguous commands, the proposed method can estimate the position of the requested object with an accuracy of 98.49% based on our test dataset. Given ambiguous language commands, we show that the accuracy of the pick up task increases by 1.94 times after incorporating the information obtained from the interaction.

Each person has a different personal space and behaves differently when another person approaches. Based on this observation, we propose a novel method to learn how to approach a person comfortably based on the person's preference while avoiding uncomfortable encounters. We propose a personal comfort field to learn each person's preference about an approaching object. A personal comfort field is based on existing theories in anthropology and personalized for each user through repeated encounters. We propose an online method to learn a personal comfort field of a user, i.e., personalized learning, based on the concept from the Gaussian process upper confidence bound and show that the proposed method has no regret asymptotically. The effectiveness of the proposed method has been extensively validated in simulation and real-world experiments. Results show that the proposed method can gradually learn the personalized approaching behavior preferred by the user as the number of encounters increases.

In this paper, we propose a generative model which learns the relationship between language and human action in order to generate a human action sequence given a sentence describing human behavior. The proposed generative model is a generative adversarial network (GAN), which is based on the sequence to sequence (SEQ2SEQ) model. Using the proposed generative network, we can synthesize various actions for a robot or a virtual agent using a text encoder recurrent neural network (RNN) and an action decoder RNN. The proposed generative network is trained from 29,770 pairs of actions and sentence annotations extracted from MSR-Video-to-Text (MSR-VTT), a large-scale video dataset. We demonstrate that the network can generate human-like actions which can be transferred to a Baxter robot, such that the robot performs an action based on a provided sentence. Results show that the proposed generative network correctly models the relationship between language and action and can generate a diverse set of actions from the same sentence.