IROS 2022 Papers

Low-light environments have posed a formidable challenge for robust unmanned aerial vehicle (UAV) tracking even with state-of-the-art (SOTA) trackers since the poten-tial image features are hard to extract under adverse light conditions. Besides, due to the low visibility, accurate online selection of the object also becomes extremely difficult for human monitors to initialize UAV tracking in ground con-trol stations. To solve these problems, this work proposes a novel enhancer, i.e., HighlightNet, to light up potential objects for both human operators and UAV trackers. By employing Transformer, HighlightNet can adjust enhancement parameters according to global features and is thus adaptive for the illumination variation. Pixel-level range mask is introduced to make HighlightNet more focused on the enhancement of the tracking object and regions without light sources. Furthermore, a soft truncation mechanism is built to prevent background noise from being mistaken for crucial features. Evaluations on image enhancement benchmarks demonstrate HighlightNet has advantages in facilitating human perception. Experiments on the public UAVDark135 benchmark show that HightlightNet is more suitable for UAV tracking tasks than other state-of-the-art (SOTA) low-light enhancers. In addition, real-world tests on a typical UAV platform verify HightlightNet's practicability and efficiency in nighttime aerial tracking-related applications. The code and demo videos are available at https://github.com/vision4robotics/HighlightNet.

Accurate control of robots at high speeds requires a control system that can take into account the kinodynamic interactions of the robot with the environment. Prior works on learning inverse kinodynamic (IKD) models of robots have shown success in capturing the complex kinodynamic effects. However, the types of control problems these approaches can be applied to are limited only to that of following pre-computed kinodynamically feasible trajectories. In this paper we present Optim-FKD, a new formulation for accurate, high-speed robot control that makes use of a learned forward kinodynamic (FKD) model and non-linear least squares optimization. Optim-FKD can be used for accurate, high speed control on any control task specifiable by a non-linear least squares objective. Optim-FKD can solve for control objectives such as path following and time-optimal control in real time, without needing access to pre-computed kinodynamically feasible trajectories. We empirically demonstrate these abilities of our approach through experiments on a scale one-tenth autonomous car. Our results show that Optim-FKD can follow desired trajectories more accurately and can find better solutions to optimal control problems than baseline approaches.

The majority of current approaches in autonomous driving rely on High-Definition (HD) maps which detail the road geometry and surrounding area. Yet, this reliance is one of the obstacles to mass deployment of autonomous vehicles due to poor scalability of such prior maps. In this paper, we tackle the problem of online road map extraction via leveraging the sensory system aboard the vehicle itself. To this end, we design a structured model where a graph representation of the road network is generated in a hierarchical fashion within a fully convolutional network. The method is able to handle complex road topology and does not require a user in the loop.

We address the problem of enabling quadrupedal robots to perform precise shooting skills in the real world using reinforcement learning. Developing algorithms to enable a legged robot to shoot a soccer ball to a given target is a challenging problem that combines robot motion control and planning into one task. To solve this problem, we need to consider the dynamics limitation and motion stability during the control of a dynamic legged robot. Moreover, we need to consider motion planning to shoot the hard-to-model deformable ball rolling on the ground with uncertain friction to a desired location. In this paper, we propose a hierarchical framework that leverages deep reinforcement learning to train (a) a robust motion control policy that can track arbitrary motions and (b) a planning policy to decide the desired kicking motion to shoot a soccer ball to a target. We deploy the proposed framework on an A1 quadrupedal robot and enable it to accurately shoot the ball to random targets in the real world.

We demonstrate the first large-scale application of model-based generative adversarial imitation learning (MGAIL) to the task of dense urban self-driving. We augment standard MGAIL using a hierarchical model to enable generalization to arbitrary goal routes, and measure performance using a closed-loop evaluation framework with simulated interactive agents. We train policies from expert trajectories collected from real vehicles driving over 100,000 miles in San Francisco, and demonstrate a steerable policy that can navigate robustly even in a zero-shot setting, generalizing to synthetic scenarios with novel goals that never occurred in real-world driving. We also demonstrate the importance of mixing closed-loop MGAIL losses with open-loop behavior cloning losses, and show our best policy approaches the performance of the expert. We evaluate our imitative model in both average and challenging scenarios, and show how it can serve as a useful prior to plan successful trajectories.

Laser-driven micro-tools are formulated by treating highly-focused laser beams as actuators, to control the tool's motion to contact then manipulate a micro object, which allows it to manipulate opaque micro objects, or large cells without causing photodamage. However, most existing laser-driven tools are limited to relatively simple tasks, such as moving and caging, and cannot carry out in-hand dexterous tasks. This is mainly because in-hand manipulation involves continuously coordinating multiple laser beams, micro-tools, and the object itself, which has high degrees of freedom (DoF) and poses up challenge for planner and controller design. This paper presents a new hierarchical formulation for the grasping and manipulation of micro objects using multiple laser-driven micro-tools. In hardware, multiple laser-driven tools are assembled to act as a robotic hand to carry out in-hand tasks (e.g., rotating); in software, a hierarchical scheme is developed to shrunken the action space and coordinate the motion of multiple tools, subject to both the parametric uncertainty in the tool and the unknown dynamic model of the object. Such a formulation provides potential for achieving robotic in-hand manipulation at a micro scale. The performance of the proposed system is validated in simulation studies under different scenarios.

Evaluating safety performance in a resource-efficient way is crucial for the development of autonomous systems. Simulation of parameterized scenarios is a popular testing strategy but parameter sweeps can be prohibitively expensive. To address this, we propose HiddenGems: a sample-efficient method for discovering the boundary between compliant and non-compliant behavior via active learning. Given a parameterized scenario, one or more compliance metrics, and a simulation oracle, HiddenGems maps the compliant and noncompliant domains of the scenario. The methodology enables critical test case identification, comparative analysis of different versions of the system under test, as well as verification of design objectives. We evaluate HiddenGems on a scenario with a jaywalker crossing in front of an autonomous vehicle and obtain compliance boundary estimates for collision, lane keep, and acceleration metrics individually and in combination, with 6 times fewer simulations than a parameter sweep. We also show how HiddenGems can be used to detect and rectify a failure mode for an unprotected turn with 86% fewer simulations.

Conversation can play an essential role in forging bonds between humans and social robots, but participants need to feel like they are being listened to, remembered, and cared about in order to effectively build rapport. In this paper, we propose a novel strategy for conducting small talk with a social robot. Our approach is known as the Tiers of Friendship. It is centered around three core design elements: 1) Persuasive content and character is provided through topic modules created by professional creative writers to ensure engaging conversational content and a compelling personality for the social robot. 2) Conversational memory is achieved by allowing topic modules to specify required information that can be learned through conversation or recalled from previous interactions and organizing topic modules into a hierarchy that enforces information requirements between topics. 3) Dynamicity in conversation is promoted through topic navigation that supports fluid transitions to topics of human interest and employs elements of random ordering to create fresh conversation experiences. In this paper, we show how the Tiers of Friendship can be used to generate conversation content for a social robot that encourages the development of rapport. We describe a working implementation of a small talk system for a social robot based on the Tiers of Friendship that combines off-the-shelf ASR and NLU components and custom robot behavior components implemented via behavior trees on ROS. Finally, in order to evaluate our approach's effectiveness, we conduct an elicitation survey that evaluates conversations in terms of perceived engagement, personality traits, and rapport expectation and discuss the implications for social robotics.

Reasoning about the future behavior of other agents is critical to safe robot navigation. The multiplicity of plausible futures is further amplified by the uncertainty inherent to agent state estimation from data, including positions, velocities, and semantic class. Forecasting methods, however, typically neglect class uncertainty, conditioning instead only on the agent's most likely class, even though perception models often return full class distributions. To exploit this information, we present HAICU, a method for heterogeneous-agent trajectory forecasting that explicitly incorporates agents' class probabilities. We additionally present PUP, a new challenging real-world autonomous driving dataset, to investigate the im-pact of Perceptual Uncertainty in Prediction. It contains chal-lenging crowded scenes with unfiltered agent class probabilities that reflect the long-tail of current state-of-the-art perception systems. We demonstrate that incorporating class probabilities in trajectory forecasting significantly improves performance in the face of uncertainty, and enables new forecasting capabilities such as counterfactual predictions.

Model Predictive Control (MPC) schemes have proven their efficiency in controlling high degree-of-freedom (DoF) complex robotic systems. However, they come at a high computational cost and an update rate of about tens of hertz. This relatively slow update rate hinders the possibility of stable haptic teleoperation of such systems since the slow feedback loops can cause instabilities and loss of transparency to the operator. This work presents a novel framework for transparent teleoperation of MPC-controlled complex robotic systems. In particular, we employ a feedback MPC approach [1] and exploit its structure to account for the operator input at a fast rate which is independent of the update rate of the MPC loop itself. We demonstrate our framework on a mobile manipulator platform and show that it significantly improves haptic teleoperation's transparency and stability. We also highlight that the proposed feedback structure is constraint satisfactory and does not violate any constraints defined in the optimal control problem. To the best of our knowledge, this work is the first realization of the bilateral teleoperation of a legged manipulator using a whole-body MPC framework.

Relocation of haptic feedback from the fingertips to the wrist has been considered as a way to enable haptic interaction with mixed reality virtual environments while leaving the fingers free for other tasks. We present a pair of wrist-worn tactile haptic devices and a virtual environment to study how various mappings between fingers and tactors affect task performance. The haptic feedback rendered to the wrist reflects the interaction forces occurring between a virtual object and virtual avatars controlled by the index finger and thumb. We performed a user study comparing four different finger-to-tactor haptic feedback mappings and one no-feedback condition as a control. We evaluated users' ability to perform a simple pick-and-place task via the metrics of task completion time, path length of the fingers and virtual cube, and magnitudes of normal and shear forces at the fingertips. We found that multiple mappings were effective, and there was a greater impact when visual cues were limited. We discuss the limitations of our approach and describe next steps toward multi-degree-of-freedom haptic rendering for wrist-worn devices to improve task performance in virtual environments.

Robotic systems capable of Dynamic Mobile Manipulation (DMM) tasks combine dynamic manipulation and locomotion and could facilitate dangerous or physically demanding labor. For instance, firefighter humanoid robots could leverage their body by leaning against collapsed building rubble to push it aside. Here we introduce a teleoperation system that targets the realization of these tasks using human's whole-body motor skills. We describe a new wheeled humanoid platform, SATYRR, and a novel hands-free teleoperation architecture using a whole-body Human Machine Interface (HMI). This system enables telelocomotion of the humanoid robot using the operator's body motion, freeing their arms for manipulation tasks. In this study we evaluate the efficacy of the proposed system on hardware, and explore the control of SATYRR using two teleoperation mappings that map the operators body pitch and yaw to the robot's velocity or acceleration. Through experiments and user feedback we showcase our preliminary findings of the pilot-system response. Results suggest that the HMI is capable of effectively telelocomoting SATYRR, that pilot preferences should dictate the appropriate motion mapping and gains, and finally that the pilot can better learn to control the system over time. This study represents a fundamental step towards the realization of combined manipulation and locomotion via teleoperation.

In most MPC-based schemes used for humanoid gait generation, simple Quadratic Programming (QP) problems are considered for real-time implementation. Since these only allow for convex constraints, the generated gait may be conservative. In this paper we focus on the non-convex reachable region of the swinging foot, also known as Kinematic Admissible Region (KAR), and the corresponding constraint. We represent an approximation of such non-convex region as the union of multiple non-overlapping convex sub-regions. By leveraging the concept of feasibility region, i.e., the subset of the state space for which a QP problem is feasible, and introducing a proper selection criterion, we are able to maintain linearity of the constraints and thus use our Intrinsically Stable Model Predictive Control (IS-MPC) scheme with a negligible additional computational load. This approach allows for a wider range of possible generated motions and is very effective when reacting to a push or avoiding an obstacle, as illustrated in dynamically simulated scenarios.

Plastic waste is a global concern that has a negative impact on the oceans and wildlife health. This paper focuses on detection of floating plastics in aerial images taken from unmanned aerial vehicles (UAVs). It proposes a new method for plastic detection in marine environments, based on SIFT descriptor and color histograms for feature extraction, as an alternative to state-of-the-art object detectors based on convolutional neural networks (CNNs), Our approach is named SURFACE: “SIFT featURes For plAstiC dEtection”. We investigate how different color-spaces and image resolutions impact the extraction of SIFT features and compare SURFACE to ResNet CNN. Also, we provide a detailed comparison with YOLO and Faster-RCNN object detection models and show that SURFACE achieves approximately the same accuracy while being faster and less memory consuming. The dataset acquired during this research will be publicly available.

Recognizing human partners is an essential social skill for building personalized and long-term human-robot interactions. However, robots deployed in complex, real-world environments have to face several challenges, such as managing unstructured interactions with multiple users, limited computational resources, and intrinsic and continuous variability of their sensory evidence. To cope with these challenges, we propose a framework to perform autonomous incremental learning for open-set face recognition suitable for unconstrained HRI scenarios. We validated the proposed framework in a real-world experiment, demonstrating its suitability to let the robot autonomously interact with multiple people while creating a labeled database of their faces across various encounters. Furthermore, we evaluated how an off-the-shelf model performed with data gathered from the HRI setting and proposed a fine-tuned model obtained with a transfer learning technique. Analyses about automatic threshold determination and rehearsal methods for memory sampling were also proposed. Our preliminary results suggest that exploiting the first-hand robot's experience could be crucial to ensure better models' performance and, therefore, could be advantageous for the acceptance and effectiveness of social robots in the long run. With this work, we aim to provide insights on continual learning approaches in the HRI field to promote autonomous and personalized solutions meaningful for real-world applications.

With the goal of efficiently computing collisionfree robot motion trajectories in dynamically changing environments, we present results of a novel method for Heuristics Informed Robot Online Path Planning (HIRO). Dividing robot environments into static and dynamic elements, we use the static part for initializing a deterministic roadmap, which provides a lower bound of the final path cost as informed heuristics for fast path-finding. These heuristics guide a search tree to explore the roadmap during runtime. The search tree examines the edges using a fuzzy collision checking concerning the dynamic environment. Finally, the heuristics tree exploits knowledge fed back from the fuzzy collision checking module and updates the lower bound for the path cost. As we demonstrate in real-world experiments, the closed-loop formed by these three components significantly accelerates the planning procedure. An additional backtracking step ensures the feasibility of the resulting paths. Experiments in simulation and the real world show that HIRO can find collisionfree paths considerably faster than baseline methods with and without prior knowledge of the environment.

This paper considers the problem of enabling the user to modify the path of a telepresence robot. The robot is capable of autonomously navigating to a goal predefined by the user, but the user might still want to modify the path, for example, to go further away from other people, or to go closer to landmarks she wants to see on the way. We propose Human-Influenced Dynamic Window Approach (HI-DWA), a shared control method aimed for telepresence robots based on Dynamic Window Approach (DWA) that allows the user to influence the control input given to the robot. To verify the proposed method, we performed a user study (N=32) in Virtual Reality (VR) to compare HI-DWA with switching between autonomous navigation and manual control for controlling a simulated telepresence robot moving in a virtual environment. Results showed that users reached their goal faster using HI-DWA controller and found it easier to use. Preference between the two methods was split equally. Qualitative analysis revealed that a major reason for the participants that preferred switching between two modes was the feeling of control. We also analyzed the effect of different input methods, joystick and gesture, on the preference and perceived workload.

Pedestrian trajectory prediction is of great importance for downstream tasks, such as autonomous driving and mobile robot navigation. Realistic models of the social interactions within the crowd is crucial for accurate pedestrian trajectory prediction. However, most existing methods do not capture group level interactions well, focusing only on pairwise interactions and neglecting group-wise interactions. In this work, we propose a hierarchical graph convolutional network, HGCN-GJS, for trajectory prediction which well leverages group level interactions within the crowd. Furthermore, we introduce a joint sampling scheme that captures co-dependencies between pedestrian trajectories during trajectory generation. Based on group information, this scheme ensures that generated trajectories within each group are consistent with each other, but enables different groups to act more independently. We demonstrate that our proposed network achieves state of the art performance on all datasets we have considered.