Dana Kulić

Mingjie Li, Ben Beck, Tharindu Rathnayake et al.

Cycling is a healthy and sustainable mode of transport. However, interactions with motor vehicles remain a key barrier to increased cycling participation. The ability to detect potentially dangerous interactions from on-bike sensing could provide important information to riders and policymakers. A key influence on rider comfort and safety is close passes, i.e., when a vehicle narrowly passes a cyclist. In this paper, we introduce a novel benchmark, called Cyc-CP, towards close pass (CP) event detection from video streams. The task is formulated into two problem categories: scene-level and instance-level. Scene-level detection ascertains the presence of a CP event within the provided video clip. Instance-level detection identifies the specific vehicle within the scene that precipitates a CP event. To address these challenges, we introduce four benchmark models, each underpinned by advanced deep-learning methodologies. For training and evaluating those models, we have developed a synthetic dataset alongside the acquisition of a real-world dataset. The benchmark evaluations reveal that the models achieve an accuracy of 88.13\% for scene-level detection and 84.60\% for instance-level detection on the real-world dataset. We envision this benchmark as a test-bed to accelerate CP detection and facilitate interaction between the fields of road safety, intelligent transportation systems and artificial intelligence. Both the benchmark datasets and detection models will be available at https://github.com/SustainableMobility/cyc-cp to facilitate experimental reproducibility and encourage more in-depth research in the field.

Thomas Beckers, Jonas Umlauft, Dana Kulić et al.

High performance tracking control can only be achieved if a good model of the dynamics is available. However, such a model is often difficult to obtain from first order physics only. In this paper, we develop a data-driven control law that ensures closed loop stability of Lagrangian systems. For this purpose, we use Gaussian Process regression for the feed-forward compensation of the unknown dynamics of the system. The gains of the feedback part are adapted based on the uncertainty of the learned model. Thus, the feedback gains are kept low as long as the learned model describes the true system sufficiently precisely. We show how to select a suitable gain adaption law that incorporates the uncertainty of the model to guarantee a globally bounded tracking error. A simulation with a robot manipulator demonstrates the efficacy of the proposed control law.

Rachel Love, Edith Law, Philip R. Cohen et al.

Conversational teachable agents offer a promising platform to support learning, both in the classroom and in remote settings. In this context, the agent takes the role of the novice, while the student takes on the role of teacher. This framing is significant for its ability to elicit the Protégé effect in the student-teacher, a pedagogical phenomenon known to increase engagement in the teaching task, and also improve cognitive outcomes. In prior work, teachable agents often take a passive role in the learning interaction, and there are few studies in which the agent and student engage in natural language dialogue during the teaching task. This work investigates the effect of teaching modality when interacting with a virtual agent, via the web-based teaching platform, the Curiosity Notebook. A method of teaching the agent by selecting sentences from source material is compared to a method paraphrasing the source material and typing text input to teach. A user study has been conducted to measure the effect teaching modality on the learning outcomes and engagement of the participants. The results indicate that teaching via paraphrasing and text input has a positive effect on learning outcomes for the material covered, and also on aspects of affective engagement. Furthermore, increased paraphrasing effort, as measured by the similarity between the source material and the material the teacher conveyed to the robot, improves learning outcomes for participants.

Lingheng Meng, Rob Gorbet, Michael Burke et al.

Deep Reinforcement Learning (DRL) has made tremendous advances in both simulated and real-world robot control tasks in recent years. This is particularly the case for tasks that can be carefully engineered with a full state representation, and which can then be formulated as a Markov Decision Process (MDP). However, applying DRL strategies designed for MDPs to novel robot control tasks can be challenging, because the available observations may be a partial representation of the state, resulting in a Partially Observable Markov Decision Process (POMDP). This paper considers three popular DRL algorithms, namely Proximal Policy Optimization (PPO), Twin Delayed Deep Deterministic Policy Gradient (TD3), and Soft Actor-Critic (SAC), invented for MDPs, and studies their performance in POMDP scenarios. While prior work has found that SAC and TD3 typically outperform PPO across a broad range of tasks that can be represented as MDPs, we show that this is not always the case, using three representative POMDP environments. Empirical studies show that this is related to multi-step bootstrapping, where multi-step immediate rewards, instead of one-step immediate reward, are used to calculate the target value estimation of an observation and action pair. We identify this by observing that the inclusion of multi-step bootstrapping in TD3 (MTD3) and SAC (MSAC) results in improved robustness in POMDP settings.

Rhys Newbury, Juyan Zhang, Tin Tran et al.

Understanding and representing the structure of 3D objects in an unsupervised manner remains a core challenge in computer vision and graphics. Most existing unsupervised keypoint methods are not designed for unconditional generative settings, restricting their use in modern 3D generative pipelines; our formulation explicitly bridges this gap. We present an unsupervised framework for learning spatially structured 3D keypoints from point cloud data. These keypoints serve as a compact and interpretable representation that conditions an Elucidated Diffusion Model (EDM) to reconstruct the full shape. The learned keypoints exhibit repeatable spatial structure across object instances and support smooth interpolation in keypoint space, indicating that they capture geometric variation. Our method achieves strong performance across diverse object categories, yielding a 6 percentage-point improvement in keypoint consistency compared to prior approaches.

Seung Chan Hong, Dana Kulić, Leimin Tian

Human-robot collaboration (HRC) can benefit from robots' abilities to interpret human emotional states. However, current emotion recognition (ER) models in HRC often fall short, particularly due to their reliance on acted datasets and single-modality inputs like facial expressions. We propose a novel vision language model (VLM)-based ER system that leverages contextual understanding to improve emotion interpretation in HRC. We first evaluate the VLM-ER system by assessing its semantic and sentiment similarity with human annotations on an existing HRC dataset. Then, in a user study with a service robot in a collaborative delivery task, we evaluate the effects of modulating the robot's behaviour based on the user's emotional state inferred by the VLM-ER system. The results show that the proposed VLM-ER system achieves higher semantic similarity and positive sentiment alignment with human annotations compared to a baseline convolutional neural network-based system. Further, participants in the user study preferred emotion-adaptive robot behaviour facilitated by the VLM-ER system.

Kal Backman, Ben Beck, Dana Kulić

While cycling offers an attractive option for sustainable transportation, many potential cyclists are discouraged from taking up cycling due to the lack of suitable and safe infrastructure. Efficiently mapping cycling infrastructure across entire cities is necessary to advance our understanding of how to provide connected networks of high-quality infrastructure. Therefore we propose a system capable of classifying available cycling infrastructure from on-bike smartphone camera data. The system receives an image sequence as input, temporally analyzing the sequence to account for sparsity of signage. The model outputs cycling infrastructure class labels defined by a hierarchical classification system. Data is collected via participant cyclists covering 7,006Km across the Greater Melbourne region that is automatically labeled via a GPS and OpenStreetMap database matching algorithm. The proposed model achieved an accuracy of 95.38%, an increase in performance of 7.55% compared to the non-temporal model. The model demonstrated robustness to extreme absence of image features where the model lost only 6.6% in accuracy after 90% of images being replaced with blank images. This work is the first to classify cycling infrastructure using only street-level imagery collected from bike-mounted mobile phone cameras, while demonstrating robustness to feature sparsity via long temporal sequence analysis.

Kerry He, Rhys Newbury, Tin Tran et al.

Occlusions caused by a robot's own body is a common problem for closed-loop control methods employed in eye-to-hand camera setups. We propose an optimization-based reactive controller that minimizes self-occlusions while achieving a desired goal pose. The approach allows coordinated control between the robot's base, arm and head by encoding the line-of-sight visibility to the target as a soft constraint along with other task-related constraints, and solving for feasible joint and base velocities. The generalizability of the approach is demonstrated in simulated and real-world experiments, on robots with fixed or mobile bases, with moving or fixed objects, and multiple objects. The experiments revealed a trade-off between occlusion rates and other task metrics. While a planning-based baseline achieved lower occlusion rates than the proposed controller, it came at the expense of highly inefficient paths and a significant drop in the task success. On the other hand, the proposed controller is shown to improve visibility to the line target object(s) without sacrificing too much from the task success and efficiency. Videos and code can be found at: rhys-newbury.github.io/projects/vmc/.

Kal Backman, Dana Kulić, Hoam Chung

Novice pilots find it difficult to operate and land unmanned aerial vehicles (UAVs), due to the complex UAV dynamics, challenges in depth perception, lack of expertise with the control interface and additional disturbances from the ground effect. Therefore we propose a shared autonomy approach to assist pilots in safely landing a UAV under conditions where depth perception is difficult and safe landing zones are limited. Our approach comprises of two modules: a perception module that encodes information onto a compressed latent representation using two RGB-D cameras and a policy module that is trained with the reinforcement learning algorithm TD3 to discern the pilot's intent and to provide control inputs that augment the user's input to safely land the UAV. The policy module is trained in simulation using a population of simulated users. Simulated users are sampled from a parametric model with four parameters, which model a pilot's tendency to conform to the assistant, proficiency, aggressiveness and speed. We conduct a user study (n = 28) where human participants were tasked with landing a physical UAV on one of several platforms under challenging viewing conditions. The assistant, trained with only simulated user data, improved task success rate from 51.4% to 98.2% despite being unaware of the human participants' goal or the structure of the environment a priori. With the proposed assistant, regardless of prior piloting experience, participants performed with a proficiency greater than the most experienced unassisted participants.

Jessy Ceha, Edith Law, Dana Kulić et al.

With advances in artificial intelligence, research is increasingly exploring the potential functions that social robots can play in education. As teachers are a critical stakeholder in the use and application of educational technologies, we conducted a study to understand teachers' perspectives on how a social robot could support a variety of learning activities in the classroom. Through interviews, robot puppeteering, and group brainstorming sessions with five elementary and middle school teachers from a local school in Canada, we take a socio-technical perspective to conceptualize possible robot functions and behaviours, and the effects they may have on the current way learning activities are designed, planned, and executed. Overall, the teachers responded positively to the idea of introducing a social robot as a technological tool for learning activities, envisioning differences in usage for teacher-robot and student-robot interactions. Further, Engeström's Activity System Model -- a framework for analyzing human needs, tasks, and outcomes -- illustrated a number of tensions associated with learning activities in the classroom. We discuss the fine-grained robot functions and behaviours conceived by teachers, and how they address the current tensions -- providing suggestions for improving the design of social robots for learning activities.

Lingheng Meng, Rob Gorbet, Dana Kulić

A promising characteristic of Deep Reinforcement Learning (DRL) is its capability to learn optimal policy in an end-to-end manner without relying on feature engineering. However, most approaches assume a fully observable state space, i.e. fully observable Markov Decision Processes (MDPs). In real-world robotics, this assumption is unpractical, because of issues such as sensor sensitivity limitations and sensor noise, and the lack of knowledge about whether the observation design is complete or not. These scenarios lead to Partially Observable MDPs (POMDPs). In this paper, we propose Long-Short-Term-Memory-based Twin Delayed Deep Deterministic Policy Gradient (LSTM-TD3) by introducing a memory component to TD3, and compare its performance with other DRL algorithms in both MDPs and POMDPs. Our results demonstrate the significant advantages of the memory component in addressing POMDPs, including the ability to handle missing and noisy observation data.

Kal Backman, Dana Kulić, Hoam Chung

Unmanned aerial vehicles (UAVs) are often used for navigating dangerous terrains, however they are difficult to pilot. Due to complex input-output mapping schemes, limited perception, the complex system dynamics and the need to maintain a safe operation distance, novice pilots experience difficulties in performing safe landings in obstacle filled environments. In this work we propose a shared autonomy approach that assists novice pilots to perform safe landings on one of several elevated platforms at a proficiency equal to or greater than experienced pilots. Our approach consists of two modules, a perceptual module and a policy module. The perceptual module compresses high dimensionality RGB-D images into a latent vector trained with a cross-modal variational auto-encoder. The policy module provides assistive control inputs trained with the reinforcement algorithm TD3. We conduct a user study (n=33) where participants land a simulated drone with and without the use of the assistant. Despite the goal platform not being known to the assistant, participants of all skill levels were able to outperform experienced participants while assisted in the task.

Vladimir Joukov, Dana Kulić

Gaussian processes regression models are an appealing machine learning method as they learn expressive non-linear models from exemplar data with minimal parameter tuning and estimate both the mean and covariance of unseen points. However, exponential computational complexity growth with the number of training samples has been a long standing challenge. During training, one has to compute and invert an $N \times N$ kernel matrix at every iteration. Regression requires computation of an $m \times N$ kernel where $N$ and $m$ are the number of training and test points respectively. In this work we show how approximating the covariance kernel using eigenvalues and functions leads to an approximate Gaussian process with significant reduction in training and regression complexity. Training with the proposed approach requires computing only a $N \times n$ eigenfunction matrix and a $n \times n$ inverse where $n$ is a selected number of eigenvalues. Furthermore, regression now only requires an $m \times n$ matrix. Finally, in a special case the hyperparameter optimization is completely independent form the number of training samples. The proposed method can regress over multiple outputs, estimate the derivative of the regressor of any order, and learn the correlations between them. The computational complexity reduction, regression capabilities, and multioutput correlation learning are demonstrated in simulation examples.

Wanxin Jin, Todd D. Murphey, Dana Kulić et al.

This paper develops the method of Continuous Pontryagin Differentiable Programming (Continuous PDP), which enables a robot to learn an objective function from a few sparsely demonstrated keyframes. The keyframes, labeled with some time stamps, are the desired task-space outputs, which a robot is expected to follow sequentially. The time stamps of the keyframes can be different from the time of the robot's actual execution. The method jointly finds an objective function and a time-warping function such that the robot's resulting trajectory sequentially follows the keyframes with minimal discrepancy loss. The Continuous PDP minimizes the discrepancy loss using projected gradient descent, by efficiently solving the gradient of the robot trajectory with respect to the unknown parameters. The method is first evaluated on a simulated robot arm and then applied to a 6-DoF quadrotor to learn an objective function for motion planning in unmodeled environments. The results show the efficiency of the method, its ability to handle time misalignment between keyframes and robot execution, and the generalization of objective learning into unseen motion conditions.

Lingheng Meng, Rob Gorbet, Dana Kulić

Multi-step (also called n-step) methods in reinforcement learning (RL) have been shown to be more efficient than the 1-step method due to faster propagation of the reward signal, both theoretically and empirically, in tasks exploiting tabular representation of the value-function. Recently, research in Deep Reinforcement Learning (DRL) also shows that multi-step methods improve learning speed and final performance in applications where the value-function and policy are represented with deep neural networks. However, there is a lack of understanding about what is actually contributing to the boost of performance. In this work, we analyze the effect of multi-step methods on alleviating the overestimation problem in DRL, where multi-step experiences are sampled from a replay buffer. Specifically building on top of Deep Deterministic Policy Gradient (DDPG), we propose Multi-step DDPG (MDDPG), where different step sizes are manually set, and its variant called Mixed Multi-step DDPG (MMDDPG) where an average over different multi-step backups is used as update target of Q-value function. Empirically, we show that both MDDPG and MMDDPG are significantly less affected by the overestimation problem than DDPG with 1-step backup, which consequently results in better final performance and learning speed. We also discuss the advantages and disadvantages of different ways to do multi-step expansion in order to reduce approximation error, and expose the tradeoff between overestimation and underestimation that underlies offline multi-step methods. Finally, we compare the computational resource needs of Twin Delayed Deep Deterministic Policy Gradient (TD3), a state-of-art algorithm proposed to address overestimation in actor-critic methods, and our proposed methods, since they show comparable final performance and learning speed.

Shray Bansal, Rhys Newbury, Wesley Chan et al.

The increasing presence of robots alongside humans, such as in human-robot teams in manufacturing, gives rise to research questions about the kind of behaviors people prefer in their robot counterparts. We term actions that support interaction by reducing future interference with others as supportive robot actions and investigate their utility in a co-located manipulation scenario. We compare two robot modes in a shared table pick-and-place task: (1) Task-oriented: the robot only takes actions to further its own task objective and (2) Supportive: the robot sometimes prefers supportive actions to task-oriented ones when they reduce future goal-conflicts. Our experiments in simulation, using a simplified human model, reveal that supportive actions reduce the interference between agents, especially in more difficult tasks, but also cause the robot to take longer to complete the task. We implemented these modes on a physical robot in a user study where a human and a robot perform object placement on a shared table. Our results show that a supportive robot was perceived as a more favorable coworker by the human and also reduced interference with the human in the more difficult of two scenarios. However, it also took longer to complete the task highlighting an interesting trade-off between task-efficiency and human-preference that needs to be considered before designing robot behavior for close-proximity manipulation scenarios.

Nils Wilde, Alexandru Blidaru, Stephen L. Smith et al.

An important challenge in human-robot interaction (HRI) is enabling non-expert users to specify complex tasks for autonomous robots. Recently, active preference learning has been applied in HRI to interactively shape a robot's behavior. We study a framework where users specify constraints on allowable robot movements on a graphical interface, yielding a robot task specification. However, users may not be able to accurately assess the impact of such constraints on the performance of a robot. Thus, we revise the specification by iteratively presenting users with alternative solutions where some constraints might be violated, and learn about the importance of the constraints from the users' choices between these alternatives. We demonstrate our framework in a user study with a material transport task in an industrial facility. We show that nearly all users accept alternative solutions and thus obtain a revised specification through the learning process, and that the revision leads to a substantial improvement in robot performance. Further, the learning process reduces the variances between the specifications from different users and, thus, makes the specifications more similar. As a result, the users whose initial specifications had the largest impact on performance benefit the most from the interactive learning.

Lingheng Meng, Daiwei Lin, Adam Francey et al.

Physical agents that can autonomously generate engaging, life-like behaviour will lead to more responsive and interesting robots and other autonomous systems. Although many advances have been made for one-to-one interactions in well controlled settings, future physical agents should be capable of interacting with humans in natural settings, including group interaction. In order to generate engaging behaviours, the autonomous system must first be able to estimate its human partners' engagement level. In this paper, we propose an approach for estimating engagement during group interaction by simultaneously taking into account active and passive interaction, i.e. occupancy, and use the measure as the reward signal within a reinforcement learning framework to learn engaging interactive behaviours. The proposed approach is implemented in an interactive sculptural system in a museum setting. We compare the learning system to a baseline using pre-scripted interactive behaviours. Analysis based on sensory data and survey data shows that adaptable behaviours within an expert-designed action space can achieve higher engagement and likeability.

Muriel Lang, Franz M. J. Pfister, Jakob Fröhner et al.

The assessment of Parkinson's disease (PD) poses a significant challenge as it is influenced by various factors which lead to a complex and fluctuating symptom manifestation. Thus, a frequent and objective PD assessment is highly valuable for effective health management of people with Parkinson's disease (PwP). Here, we propose a method for monitoring PwP by stochastically modeling the relationships between their wrist movements during unscripted daily activities and corresponding annotations about clinical displays of movement abnormalities. We approach the estimation of PD motor signs by independently modeling and hierarchically stacking Gaussian process models for three classes of commonly observed movement abnormalities in PwP including tremor, (non-tremulous) bradykinesia, and (non-tremulous) dyskinesia. We use clinically adopted severity measures as annotations for training the models, thus allowing our multi-layer Gaussian process prediction models to estimate not only their presence but also their severities. The experimental validation of our approach demonstrates strong agreement of the model predictions with these PD annotations. Our results show the proposed method produces promising results in objective monitoring of movement abnormalities of PD in the presence of arbitrary and unknown voluntary motions, and makes an important step towards continuous monitoring of PD in the home environment.

Terry Taewoong Um, Franz Michael Josef Pfister, Daniel Christian Pichler et al.

Parkinson's Disease (PD) is characterized by disorders in motor function such as freezing of gait, rest tremor, rigidity, and slowed and hyposcaled movements. Medication with dopaminergic medication may alleviate those motor symptoms, however, side-effects may include uncontrolled movements, known as dyskinesia. In this paper, an automatic PD motor-state assessment in free-living conditions is proposed using an accelerometer in a wrist-worn wearable sensor. In particular, an ensemble of convolutional neural networks (CNNs) is applied to capture the large variability of daily-living activities and overcome the dissimilarity between training and test patients due to the inter-patient variability. In addition, class activation map (CAM), a visualization technique for CNNs, is applied for providing an interpretation of the results.

Thomas Beckers, Dana Kulić, Sandra Hirche

Perfect tracking control for real-world Euler-Lagrange systems is challenging due to uncertainties in the system model and external disturbances. The magnitude of the tracking error can be reduced either by increasing the feedback gains or improving the model of the system. The latter is clearly preferable as it allows to maintain good tracking performance at low feedback gains. However, accurate models are often difficult to obtain. In this article, we address the problem of stable high-performance tracking control for unknown Euler-Lagrange systems. In particular, we employ Gaussian Process regression to obtain a data-driven model that is used for the feed-forward compensation of unknown dynamics of the system. The model fidelity is used to adapt the feedback gains allowing low feedback gains in state space regions of high model confidence. The proposed control law guarantees a globally bounded tracking error with a specific probability. Simulation studies demonstrate the superiority over state of the art tracking control approaches.

Wanxin Jin, Dana Kulić, Shaoshuai Mou et al.

This article develops a methodology that enables learning an objective function of an optimal control system from incomplete trajectory observations. The objective function is assumed to be a weighted sum of features (or basis functions) with unknown weights, and the observed data is a segment of a trajectory of system states and inputs. The proposed technique introduces the concept of the recovery matrix to establish the relationship between any available segment of the trajectory and the weights of given candidate features. The rank of the recovery matrix indicates whether a subset of relevant features can be found among the candidate features and the corresponding weights can be learned from the segment data. The recovery matrix can be obtained iteratively and its rank non-decreasing property shows that additional observations may contribute to the objective learning. Based on the recovery matrix, a method for using incomplete trajectory observations to learn the weights of selected features is established, and an incremental inverse optimal control algorithm is developed by automatically finding the minimal required observation. The effectiveness of the proposed method is demonstrated on a linear quadratic regulator system and a simulated robot manipulator.

Terry Taewoong Um, Franz Michael Josef Pfister, Daniel Pichler et al.

While convolutional neural networks (CNNs) have been successfully applied to many challenging classification applications, they typically require large datasets for training. When the availability of labeled data is limited, data augmentation is a critical preprocessing step for CNNs. However, data augmentation for wearable sensor data has not been deeply investigated yet. In this paper, various data augmentation methods for wearable sensor data are proposed. The proposed methods and CNNs are applied to the classification of the motor state of Parkinson's Disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability. Appropriate augmentation improves the classification performance from 77.54\% to 86.88\%.

Terry Taewoong Um, Vahid Babakeshizadeh, Dana Kulić

The ability to accurately identify human activities is essential for developing automatic rehabilitation and sports training systems. In this paper, large-scale exercise motion data obtained from a forearm-worn wearable sensor are classified with a convolutional neural network (CNN). Time-series data consisting of accelerometer and orientation measurements are formatted as images, allowing the CNN to automatically extract discriminative features. A comparative study on the effects of image formatting and different CNN architectures is also presented. The best performing configuration classifies 50 gym exercises with 92.1% accuracy.

Rajan Gill, Dana Kulić, Christopher Nielsen

Path following controllers make the output of a control system approach and traverse a pre-specified path with no apriori time parametrization. In this paper we present a method for path following control design applicable to framed curves generated by splines in the workspace of kinematically redundant mechanical systems. The class of admissible paths includes self-intersecting curves. Kinematic redundancies are resolved by designing controllers that solve a suitably defined constrained quadratic optimization problem. By employing partial feedback linearization, the proposed path following controllers have a clear physical meaning. The approach is experimentally verified on a 4-degree-of-freedom (4-DOF) manipulator with a combination of revolute and linear actuated links and significant model uncertainty.