Siwei Zhang

Autonomous robotic systems will be the future of planetary exploration missions. For autonomous robotic exploration, reliable pose estimation is required. This can be provided by cooperative radio localization, where radio signals are exchanged among the robots and other mission objects. Range and direction information is obtained by measuring the signal round-trip time (RTT) and direction-of-arrival (DoA), which enables position and orientation estimation of the robots. For the first time, we have demonstrated cooperative radio localization within a space-analogue exploration mission with two robotic rovers on the volcano Mt Etna. With this paper, we share our main lessons learned based on a thorough evaluation of measurement data. Thereby we identify estimation biases as the main error source. We then show how to estimate and compensate these biases during the mission by simultaneous localization and calibration (SLAC). We further investigate the impact of the radio signals reflected on the ground and on mission objects on the ranging accuracy. Then, we demonstrate the benefit of cooperation and the feasibility of single-link localization. In addition, we share tangible ranging and DoA estimation error models based on measurements in a realistic environment.

A swarm of robots offers significant advantages over a single robot, enabling faster exploration of larger areas and enhanced robustness against single-point failures. Accurate relative positioning is critical for executing collaborative missions without collisions. When Visual Simultaneous Localization and Mapping (VSLAM) is employed to estimate each robot’s poses, the inter-agent loop closing method is commonly used to improve relative positioning accuracy by refining pose estimates and merging local maps based on shared feature points. However, this approach demands considerable computational and communication resources. In this paper, we introduce Collaborative SLAM using Visual Odometry and Ranges (CoVOR-SLAM) to address these challenges. CoVOR-SLAM significantly reduces both the amount of data transmitted between robots and their computational loads, as it requires only pose estimates, covariances, and range measurements instead of feature or map points for inter-agent loop detection. The necessary range measurements can be derived from pilot signals within the communication system, eliminating the need for complex additional infrastructure. We evaluated CoVOR-SLAM using real images and ultra-wideband range data collected from two rovers, as well as in a larger multi-agent setup utilizing public image datasets and realistic simulations. The results demonstrate that CoVOR-SLAM accurately estimates robot poses while requiring considerably less computational power and communication capacity than inter-agent loop closing techniques.

Estimating the distance between the source of a diffusive process and a receiver has a variety of applications, ranging from gas source localization at the macro-scale to molecular communication at the micro-scale. Distance information can be extracted from features of the observed particle concentration, e.g. its peak. This paper derives the Cramér-Rao lower bound (CRB) for distance estimation given the advection-diffusion model for absorbing receivers, which is the fundamental limit of any distance estimator. Furthermore, CRBs are obtained for estimators using only information about the observed peak. A maximum-likelihood estimator using the entire signal and two estimators based on peak detection are deduced. The derived CRBs are used to study the effect of channel parameters on the estimation performance. Finally, the performance of the proposed estimators is verified by comparing the root mean squared errors with their theoretical bounds in a simulation, and preliminary experimental results are presented.

Autonomous robotic swarms are envisioned for a variety of applications—for example, space exploration, search and rescue, and disaster management. Important features of a robotic swarm include its ability to share information within the network, to sense spatio-temporal processes such as gas distributions, and to collaboratively enhance its navigation. In environments without infrastructure, the swarm elements can cooperatively estimate their position, e.g., based on the time of flight of exchanged radio signals. Cooperative positioning performance depends on the radio propagation environment. Free-space path loss is commonly used for performance assessment, which is an optimistic assumption. In this work, we investigate the limits to cooperative positioning and ranging based on the time of flight of radio signals over the more realistic two-ray ground reflection channel. We show that we obtain a ranging bias caused by the radio signal component reflected from the ground, and that the ranging error becomes bias-limited. In the positioning domain, we investigate how the ranging bias affects the cooperative positioning performance. As a result, we gain in cooperation, but the achievable positioning performance is significantly worsened by the ranging bias. As a conclusion, the two-ray ground reflection model should be considered to obtain realistic cooperative positioning limits.

The synthesis of human grasping has numerous applications including AR/VR, video games and robotics. While methods have been proposed to generate realistic hand-object interaction for object grasping and manipulation, these typically only consider interacting hand alone. Our goal is to synthesize whole-body grasping motions. Starting from an arbitrary initial pose, we aim to generate diverse and natural whole-body human motions to approach and grasp a target object in 3D space. This task is challenging as it requires modeling both whole-body dynamics and dexterous finger movements. To this end, we propose SAGA (StochAstic whole-body Grasping with contAct), a framework which consists of two key components: (a) Static whole-body grasping pose generation. Specifically, we propose a multi-task generative model, to jointly learn static whole-body grasping poses and human-object contacts. (b) Grasping motion infilling. Given an initial pose and the generated whole-body grasping pose as the start and end of the motion respectively, we design a novel contact-aware generative motion infilling module to generate a diverse set of grasp-oriented motions. We demonstrate the effectiveness of our method, which is a novel generative framework to synthesize realistic and expressive whole-body motions that approach and grasp randomly placed unseen objects. Code and models are available at https://jiahaoplus.github.io/SAGA/saga.html.

The DoA of radio waves is used for many applications, e.g. the localization of autonomous robots and smart vehicles. Estimating the DoA is possible with a multiport antenna, e.g. an antenna array or a multi-mode antenna (MMA). In practice, DoA estimation performance decisively depends on accurate knowledge of the antenna response, which makes antenna calibration vital. As the antenna surroundings influence its response, it is necessary to measure the entire device with installed antenna to obtain the installed antenna response. Antenna calibration is often done in a dedicated measurement chamber, which can be inconvenient and costly, especially for larger devices. Thus, auto- and in-situ calibration methods aim at making antenna calibration in a measurement chamber redundant. However, existing auto- and in-situ calibration methods are restricted to certain antenna types and certain calibrations. In this paper, we propose a Bayesian in-situ calibration algorithm based on a maximum a posteriori (MAP) estimator, which is suitable for arbitrary multiport antennas. The algorithm uses received signals from a transmitter, noisy external DoA observations, takes multipath propagation into account and does not require synchronization. Furthermore, we take an estimation theoretic perspective and provide an in-depth theoretical discussion of in-situ antenna calibration in unknown propagation conditions based on Bayesian information and the Bayesian Cramér-Rao bound (BCRB). Extensive simulations show that the proposed algorithm operates close to the BCRB and the achieved DoA estimation performance asymptotically approaches the case of a perfectly known antenna response. Finally, we provide an experimental validation, where we calibrate the antenna on a robotic rover and evaluate the DoA estimation performance using measurement data. With the proposed in-situ antenna calibration algorithm, DoA estimation performance is considerably improved compared to using an antenna response obtained by simulation or in a measurement chamber.

Cooperative radio localization and navigation systems can be used in scenarios where the reception of global navigation satellite system (GNSS) signals is not possible or impaired. While the benefit of cooperation has been highlighted by many papers, calibration is not widely considered, but equally important in practice. Utilizing the signal propagation time requires group delay or ranging bias calibration and estimating the directions-of-arrival (DoAs) requires antenna response calibration. Often, calibration parameters are determined only once before operation. However, the calibration parameters are influenced by e.g. changing temperatures of radio frequency (RF) components or changing surroundings of antennas. To cope with that, we derive a cooperative simultaneous localization and calibration (SLAC) algorithm based on Bayesian filtering, which estimates antenna responses and ranging biases simultaneously with positions and orientations. By simulations, we show that the calibration parameters can be estimated during operation without additional sensors. We further proof practical applicability of SLAC by evaluating measurement data from robotic rovers. With SLAC, both ranging and DoAs estimation performance is improved, resulting in better position and orientation estimation accuracy. SLAC is thus able to provide reliable calibration and to mitigate model mismatch. Finally, we discuss open research questions and possible extensions of SLAC.

In multipath assisted positioning, multipath components (MPCs) are regarded as line-of-sight (LoS) signals from virtual transmitters. Instead of trying to mitigate the influence of MPCs, the spatial information contained in MPCs is exploited for localization. The locations of the physical and virtual transmitters are in general unknown but can be estimated with simultaneous localization and mapping (SLAM). Recently, a multipath assisted positioning algorithm named Channel-SLAM for terrestrial radio signals has been introduced. It simultaneously tracks the position of a receiver and maps the locations of physical and virtual radio transmitters. Maps of estimated transmitter locations can be augmented by additional information. Within this paper, we propose to extend the Channel-SLAM algorithm by mapping information about the visibility of transmitters. A physical or virtual transmitter is visible, if its signal is received in a LoS condition. We derive a novel particle filter for Channel-SLAM that estimates and exploits visibility information on transmitters in addition to their locations. We show by means of simulations in an indoor scenario that our novel particle filter improves the positioning performance of Channel-SLAM considerably.

Teams of mobile robots will play a crucial role in future missions to explore the surfaces of extraterrestrial bodies. Setting up infrastructure and taking scientific samples are expensive tasks when operating in distant, challenging, and unknown environments. In contrast to current single-robot space missions, future heterogeneous robotic teams will increase efficiency via enhanced autonomy and parallelization, improve robustness via functional redundancy, as well as benefit from complementary capabilities of the individual robots. In this letter, we present our heterogeneous robotic team, consisting of flying and driving robots that we plan to deploy on scientific sampling demonstration missions at a Moon-analogue site on Mt. Etna, Sicily, Italy in 2021 as part of the ARCHES project. We describe the robots’ individual capabilities and their roles in two mission scenarios. We then present components and experiments on important tasks therein: automated task planning, high-level mission control, spectral rock analysis, radio-based localization, collaborative multi-robot 6D SLAM in Moon-analogue and Mars-like scenarios, and demonstrations of autonomous sample return.

A multitude of autonomous robotic platforms collectively organized as a swarm attracts increasing attention for remote sensing and exploration tasks. A navigation system is essential for the swarm to collectively localize itself as well as external sources. In this article, we propose a self-aware swarm navigation system that is conscious of the causality between its position and the localization uncertainty. This knowledge allows the swarm to move in a way to not only account for external mission objectives but also enhance position information. Position information for classical navigation systems has already been studied with the Fisher information (FI) and Bayesian information (BI) theories. We show how to extend these theories to a self-aware swarm navigation system, particularly emphasizing the collective performance. In this respect, fundamental limits and geometric interpretations of localization with generic observation models are discussed. We further propose a general concept of FI and BI based information seeking swarm control. The weighted position Cramér-Rao bound (CRB) and posterior CRB (PCRB) are employed flexibly as either a control cost function or constraints according to different mission criteria. As a result, the swarm actively adapts its position to enrich position information with different emerging collective behaviors. The proposed concept is illustrated by a case study of a swarm mission for gas exploration on Mars.