Dylan P. Losey

Robert Ramirez Sanchez, Daniel J. Evans, Dylan P. Losey et al.

Humans often take longer to demonstrate a task than a robot would need to execute it. Rather than learning to replicate the demonstration at the same pace, many industrial and practical applications require robots to perform tasks as quickly as possible. In this paper, we investigate several hypotheses for learning policies that operate faster-than-demonstrations. Our experiments show that the most effective strategy is to downsample recorded demonstrations and train the robot's policy on this accelerated data. However, uniformly downsampling an entire trajectory can be problematic. Some parts of a task can be safely sped up (e.g., unconstrained motion), while others demand slower, more precise motion (e.g., object interactions or fine manipulation). To address this challenge, we introduce VOLT, a vision-and-language trajectory segmentation method that reasons over video demonstrations, and leverages contextual cues to determine when acceleration is appropriate and when careful precision is required. VOLT identifies segments where slow, deliberate motion is necessary, then selectively downsamples the remaining segments. The resulting reformatted trajectories can be used with standard imitation learning approaches, such as diffusion policies. Our results highlight that segmentation quality is critical -- baseline methods often misidentify when acceleration is possible, leading to overly cautious or unreliable policies. Compared to state-of-the-art alternatives, VOLT allows robots to execute tasks faster while maintaining strong performance.

Yinlong Dai, Benjamin A. Christie, Daniel J. Evans et al.

Enabling robots to perform novel manipulation tasks from natural language instructions remains a fundamental challenge in robotics, despite significant progress in generalized problem solving with foundational models. Large vision and language models (VLMs) are capable of processing high-dimensional input data for visual scene and language understanding, as well as decomposing tasks into a sequence of logical steps; however, they struggle to ground those steps in embodied robot motion. On the other hand, robotics foundation models output action commands, but require in-domain fine-tuning or experience before they are able to perform novel tasks successfully. At its core, there still remains the fundamental challenge of connecting abstract task reasoning with low-level motion control. To address this disconnect, we propose Language Movement Primitives (LMPs), a framework that grounds VLM reasoning in Dynamic Movement Primitive (DMP) parameterization. Our key insight is that DMPs provide a small number of interpretable parameters, and VLMs can set these parameters to specify diverse, continuous, and stable trajectories. Put another way: VLMs can reason over free-form natural language task descriptions, and semantically ground their desired motions into DMPs -- bridging the gap between high-level task reasoning and low-level position and velocity control. Building on this combination of VLMs and DMPs, we formulate our LMP pipeline for zero-shot robot manipulation that effectively completes tabletop manipulation problems by generating a sequence of DMP motions. Across 31 real-world manipulation tasks, we show that LMP achieves 65% task success as compared to 35% for the best performing baseline. See videos at our website: https://collab.me.vt.edu/lmp

Sagar Parekh, Soheil Habibian, Dylan P. Losey

When robots interact with human partners, often these partners change their behavior in response to the robot. On the one hand this is challenging because the robot must learn to coordinate with a dynamic partner. But on the other hand -- if the robot understands these dynamics -- it can harness its own behavior, influence the human, and guide the team towards effective collaboration. Prior research enables robots to learn to influence other robots or simulated agents. In this paper we extend these learning approaches to now influence humans. What makes humans especially hard to influence is that -- not only do humans react to the robot -- but the way a single user reacts to the robot may change over time, and different humans will respond to the same robot behavior in different ways. We therefore propose a robust approach that learns to influence changing partner dynamics. Our method first trains with a set of partners across repeated interactions, and learns to predict the current partner's behavior based on the previous states, actions, and rewards. Next, we rapidly adapt to new partners by sampling trajectories the robot learned with the original partners, and then leveraging those existing behaviors to influence the new partner dynamics. We compare our resulting algorithm to state-of-the-art baselines across simulated environments and a user study where the robot and participants collaborate to build towers. We find that our approach outperforms the alternatives, even when the partner follows new or unexpected dynamics. Videos of the user study are available here: https://youtu.be/lYsWM8An18g

Sagar Parekh, Dylan P. Losey

When robots interact with humans in homes, roads, or factories the human's behavior often changes in response to the robot. Non-stationary humans are challenging for robot learners: actions the robot has learned to coordinate with the original human may fail after the human adapts to the robot. In this paper we introduce an algorithmic formalism that enables robots (i.e., ego agents) to co-adapt alongside dynamic humans (i.e., other agents) using only the robot's low-level states, actions, and rewards. A core challenge is that humans not only react to the robot's behavior, but the way in which humans react inevitably changes both over time and between users. To deal with this challenge, our insight is that -- instead of building an exact model of the human -- robots can learn and reason over high-level representations of the human's policy and policy dynamics. Applying this insight we develop RILI: Robustly Influencing Latent Intent. RILI first embeds low-level robot observations into predictions of the human's latent strategy and strategy dynamics. Next, RILI harnesses these predictions to select actions that influence the adaptive human towards advantageous, high reward behaviors over repeated interactions. We demonstrate that -- given RILI's measured performance with users sampled from an underlying distribution -- we can probabilistically bound RILI's expected performance across new humans sampled from the same distribution. Our simulated experiments compare RILI to state-of-the-art representation and reinforcement learning baselines, and show that RILI better learns to coordinate with imperfect, noisy, and time-varying agents. Finally, we conduct two user studies where RILI co-adapts alongside actual humans in a game of tag and a tower-building task. See videos of our user studies here: https://youtu.be/WYGO5amDXbQ

Robert Ramirez Sanchez, Heramb Nemlekar, Shahabedin Sagheb et al.

Imitation learning enables robots to learn new tasks from human examples. One fundamental limitation while learning from humans is causal confusion. Causal confusion occurs when the robot's observations include both task-relevant and extraneous information: for instance, a robot's camera might see not only the intended goal, but also clutter and changes in lighting within its environment. Because the robot does not know which aspects of its observations are important a priori, it often misinterprets the human's examples and fails to learn the desired task. To address this issue, we highlight that -- while the robot learner may not know what to focus on -- the human teacher does. In this paper we propose that the human proactively marks key parts of their task with small, lightweight beacons. Under our framework (RECON) the human attaches these beacons to task-relevant objects before providing demonstrations: as the human shows examples of the task, beacons track the position of marked objects. We then harness this offline beacon data to train a task-relevant state embedding. Specifically, we embed the robot's observations to a latent state that is correlated with the measured beacon readings: in practice, this causes the robot to autonomously filter out extraneous observations and make decisions based on features learned from the beacon data. Our simulations and a real robot experiment suggest that this framework for human-placed beacons mitigates causal confusion. Indeed, we find that using RECON significantly reduces the number of demonstrations needed to convey the task, lowering the overall time required for human teaching. See videos here: https://youtu.be/oy85xJvtLSU

Sagar Parekh, Lauren Bramblett, Nicola Bezzo et al.

Humans interacting with robots often form predictions of what the robot will do next. For instance, based on the recent behavior of an autonomous car, a nearby human driver might predict that the car is going to remain in the same lane. It is important for the robot to understand the human's prediction for safe and seamless interaction: e.g., if the autonomous car knows the human thinks it is not merging -- but the autonomous car actually intends to merge -- then the car can adjust its behavior to prevent an accident. Prior works typically assume that humans make precise predictions of robot behavior. However, recent research on human-human prediction suggests the opposite: humans tend to approximate other agents by predicting their high-level behaviors. We apply this finding to develop a second-order theory of mind approach that enables robots to estimate how humans predict they will behave. To extract these high-level predictions directly from data, we embed the recent human and robot trajectories into a discrete latent space. Each element of this latent space captures a different type of behavior (e.g., merging in front of the human, remaining in the same lane) and decodes into a vector field across the state space that is consistent with the underlying behavior type. We hypothesize that our resulting high-level and course predictions of robot behavior will correspond to actual human predictions. We provide initial evidence in support of this hypothesis through proof-of-concept simulations, testing our method's predictions against those of real users, and experiments on a real-world interactive driving dataset.

Yinlong Dai, Robert Ramirez Sanchez, Ryan Jeronimus et al.

Today's robots attempt to learn new tasks by imitating human examples. These robots watch the human complete the task, and then try to match the actions taken by the human expert. However, this standard approach to visual imitation learning is fundamentally limited: the robot observes what the human does, but not why the human chooses those behaviors. Without understanding which features of the system or environment factor into the human's decisions, robot learners often misinterpret the human's examples. In practice, this results in causal confusion, inefficient learning, and robot policies that fail when the environment changes. We therefore propose a shift in perspective: instead of asking human teachers just to show what actions the robot should take, we also enable humans to intuitively indicate why they made those decisions. Under our paradigm human teachers attach markers to task-relevant objects and use natural language prompts to describe their state representation. Our proposed algorithm, CIVIL, leverages this augmented demonstration data to filter the robot's visual observations and extract a feature representation that aligns with the human teacher. CIVIL then applies these causal features to train a transformer-based policy that -- when tested on the robot -- is able to emulate human behaviors without being confused by visual distractors or irrelevant items. Our simulations and real-world experiments demonstrate that robots trained with CIVIL learn both what actions to take and why to take those actions, resulting in better performance than state-of-the-art baselines. From the human's perspective, our user study reveals that this new training paradigm actually reduces the total time required for the robot to learn the task, and also improves the robot's performance in previously unseen scenarios. See videos at our project website: https://civil2025.github.io

Yinlong Dai, Andre Keyser, Dylan P. Losey

Imitation learning (IL) has proven effective across a wide range of manipulation tasks. However, IL policies often struggle when faced with out-of-distribution observations; for instance, when the target object is in a previously unseen position or occluded by other objects. In these cases, extensive demonstrations are needed for current IL methods to reach robust and generalizable behaviors. But when humans are faced with these sorts of atypical initial states, we often rearrange the environment for more favorable task execution. For example, a person might rotate a coffee cup so that it is easier to grasp the handle, or push a box out of the way so they can directly grasp their target object. In this work we seek to equip robot learners with the same capability: enabling robots to prepare the environment before executing their given policy. We propose ReSET, an algorithm that takes initial states -- which are outside the policy's distribution -- and autonomously modifies object poses so that the restructured scene is similar to training data. Theoretically, we show that this two step process (rearranging the environment before rolling out the given policy) reduces the generalization gap. Practically, our ReSET algorithm combines action-agnostic human videos with task-agnostic teleoperation data to i) decide when to modify the scene, ii) predict what simplifying actions a human would take, and iii) map those predictions into robot action primitives. Comparisons with diffusion policies, VLAs, and other baselines show that using ReSET to prepare the environment enables more robust task execution with equal amounts of total training data. See videos at our project website: https://reset2025paper.github.io/

Benjamin A. Christie, Dylan P. Losey

Safety is critical during human-robot interaction. But -- because people are inherently unpredictable -- it is often difficult for robots to plan safe behaviors. Instead of relying on our ability to anticipate humans, here we identify robot policies that are robust to unexpected human decisions. We achieve this by formulating human-robot interaction as a zero-sum game, where (in the worst case) the human's actions directly conflict with the robot's objective. Solving for the Nash Equilibrium of this game provides robot policies that maximize safety and performance across a wide range of human actions. Existing approaches attempt to find these optimal policies by leveraging Hamilton-Jacobi analysis (which is intractable) or linear-quadratic approximations (which are inexact). By contrast, in this work we propose a computationally efficient and theoretically justified method that converges towards the Nash Equilibrium policy. Our approach (which we call MCLQ) leverages linear-quadratic games to obtain an initial guess at safe robot behavior, and then iteratively refines that guess with a Monte Carlo search. Not only does MCLQ provide real-time safety adjustments, but it also enables the designer to tune how conservative the robot is -- preventing the system from focusing on unrealistic human behaviors. Our simulations and user study suggest that this approach advances safety in terms of both computation time and expected performance. See videos of our experiments here: https://youtu.be/KJuHeiWVuWY.

Ananth Jonnavittula, Dylan P. Losey

When humans control robot arms these robots often need to infer the human's desired task. Prior research on assistive teleoperation and shared autonomy explores how robots can determine the desired task based on the human's joystick inputs. In order to perform this inference the robot relies on an internal mapping between joystick inputs and discrete tasks: e.g., pressing the joystick left indicates that the human wants a plate, while pressing the joystick right indicates a cup. This approach works well after the human understands how the robot interprets their inputs -- but inexperienced users still have to learn these mappings through trial and error! Here we recognize that the robot's mapping between tasks and inputs is a convention. There are multiple, equally efficient conventions that the robot could use: rather than passively waiting for the human, we introduce a shared autonomy approach where the robot actively reveals its chosen convention. Across repeated interactions the robot intervenes and exaggerates the arm's motion to demonstrate more efficient inputs while also assisting for the current task. We compare this approach to a state-of-the-art baseline -- where users must identify the convention by themselves -- as well as written instructions. Our user study results indicate that modifying the robot's behavior to reveal its convention outperforms the baselines and reduces the amount of time that humans spend controlling the robot. See videos of our user study here: https://youtu.be/jROTVOp469I

Antonio Alvarez Valdivia, Ritish Shailly, Naman Seth et al.

Physical interaction between humans and robots can help robots learn to perform complex tasks. The robot arm gains information by observing how the human kinesthetically guides it throughout the task. While prior works focus on how the robot learns, it is equally important that this learning is transparent to the human teacher. Visual displays that show the robot's uncertainty can potentially communicate this information; however, we hypothesize that visual feedback mechanisms miss out on the physical connection between the human and robot. In this work we present a soft haptic display that wraps around and conforms to the surface of a robot arm, adding a haptic signal at an existing point of contact without significantly affecting the interaction. We demonstrate how soft actuation creates a salient haptic signal while still allowing flexibility in device mounting. Using a psychophysics experiment, we show that users can accurately distinguish inflation levels of the wrapped display with an average Weber fraction of 11.4%. When we place the wrapped display around the arm of a robotic manipulator, users are able to interpret and leverage the haptic signal in sample robot learning tasks, improving identification of areas where the robot needs more training and enabling the user to provide better demonstrations. See videos of our device and user studies here: https://youtu.be/tX-2Tqeb9Nw

Shaunak A. Mehta, Sagar Parekh, Dylan P. Losey

We can make it easier for disabled users to control assistive robots by mapping the user's low-dimensional joystick inputs to high-dimensional, complex actions. Prior works learn these mappings from human demonstrations: a non-disabled human either teleoperates or kinesthetically guides the robot arm through a variety of motions, and the robot learns to reproduce the demonstrated behaviors. But this framework is often impractical - disabled users will not always have access to external demonstrations! Here we instead learn diverse teleoperation mappings without either human demonstrations or pre-defined tasks. Under our unsupervised approach the robot first optimizes for object state entropy: i.e., the robot autonomously learns to push, pull, open, close, or otherwise change the state of nearby objects. We then embed these diverse, object-oriented behaviors into a latent space for real-time control: now pressing the joystick causes the robot to perform dexterous motions like pushing or opening. We experimentally show that - with a best-case human operator - our unsupervised approach actually outperforms the teleoperation mappings learned from human demonstrations, particularly if those demonstrations are noisy or imperfect. But our user study results were less clear-cut: although our approach enabled participants to complete tasks more quickly and with fewer changes of direction, users were confused when the unsupervised robot learned unexpected behaviors. See videos of the user study here: https://youtu.be/BkqHQjsUKDg

James F. Mullen, Josh Mosier, Sounak Chakrabarti et al.

Robots learn as they interact with humans. Consider a human teleoperating an assistive robot arm: as the human guides and corrects the arm's motion, the robot gathers information about the human's desired task. But how does the human know what their robot has inferred? Today's approaches often focus on conveying intent: for instance, upon legible motions or gestures to indicate what the robot is planning. However, closing the loop on robot inference requires more than just revealing the robot's current policy: the robot should also display the alternatives it thinks are likely, and prompt the human teacher when additional guidance is necessary. In this paper we propose a multimodal approach for communicating robot inference that combines both passive and active feedback. Specifically, we leverage information-rich augmented reality to passively visualize what the robot has inferred, and attention-grabbing haptic wristbands to actively prompt and direct the human's teaching. We apply our system to shared autonomy tasks where the robot must infer the human's goal in real-time. Within this context, we integrate passive and active modalities into a single algorithmic framework that determines when and which type of feedback to provide. Combining both passive and active feedback experimentally outperforms single modality baselines; during an in-person user study, we demonstrate that our integrated approach increases how efficiently humans teach the robot while simultaneously decreasing the amount of time humans spend interacting with the robot. Videos here: https://youtu.be/swq_u4iIP-g

Ananth Jonnavittula, Dylan P. Losey

Wheelchair-mounted robotic arms (and other assistive robots) should help their users perform everyday tasks. One way robots can provide this assistance is shared autonomy. Within shared autonomy, both the human and robot maintain control over the robot's motion: as the robot becomes confident it understands what the human wants, it increasingly intervenes to automate the task. But how does the robot know what tasks the human may want to perform in the first place? Today's shared autonomy approaches often rely on prior knowledge: for example, the robot must know the set of possible human goals a priori. In the long-term, however, this prior knowledge will inevitably break down -- sooner or later the human will reach for a goal that the robot did not expect. In this paper we propose a learning approach to shared autonomy that takes advantage of repeated interactions. Learning to assist humans would be impossible if they performed completely different tasks at every interaction: but our insight is that users living with physical disabilities repeat important tasks on a daily basis (e.g., opening the fridge, making coffee, and having dinner). We introduce an algorithm that exploits these repeated interactions to recognize the human's task, replicate similar demonstrations, and return control when unsure. As the human repeatedly works with this robot, our approach continually learns to assist tasks that were never specified beforehand: these tasks include both discrete goals (e.g., reaching a cup) and continuous skills (e.g., opening a drawer). Across simulations and an in-person user study, we demonstrate that robots leveraging our approach match existing shared autonomy methods for known goals, and outperform imitation learning baselines on new tasks. See videos here: https://youtu.be/Plh4t3wQeIA

Dylan P. Losey, Hong Jun Jeon, Mengxi Li et al.

Assistive robot arms enable people with disabilities to conduct everyday tasks on their own. These arms are dexterous and high-dimensional; however, the interfaces people must use to control their robots are low-dimensional. Consider teleoperating a 7-DoF robot arm with a 2-DoF joystick. The robot is helping you eat dinner, and currently you want to cut a piece of tofu. Today's robots assume a pre-defined mapping between joystick inputs and robot actions: in one mode the joystick controls the robot's motion in the x-y plane, in another mode the joystick controls the robot's z-yaw motion, and so on. But this mapping misses out on the task you are trying to perform! Ideally, one joystick axis should control how the robot stabs the tofu and the other axis should control different cutting motions. Our insight is that we can achieve intuitive, user-friendly control of assistive robots by embedding the robot's high-dimensional actions into low-dimensional and human-controllable latent actions. We divide this process into three parts. First, we explore models for learning latent actions from offline task demonstrations, and formalize the properties that latent actions should satisfy. Next, we combine learned latent actions with autonomous robot assistance to help the user reach and maintain their high-level goals. Finally, we learn a personalized alignment model between joystick inputs and latent actions. We evaluate our resulting approach in four user studies where non-disabled participants reach marshmallows, cook apple pie, cut tofu, and assemble dessert. We then test our approach with two disabled adults who leverage assistive devices on a daily basis.

Dylan P. Losey, Andrea Bajcsy, Marcia K. O'Malley et al.

When a robot performs a task next to a human, physical interaction is inevitable: the human might push, pull, twist, or guide the robot. The state-of-the-art treats these interactions as disturbances that the robot should reject or avoid. At best, these robots respond safely while the human interacts; but after the human lets go, these robots simply return to their original behavior. We recognize that physical human-robot interaction (pHRI) is often intentional -- the human intervenes on purpose because the robot is not doing the task correctly. In this paper, we argue that when pHRI is intentional it is also informative: the robot can leverage interactions to learn how it should complete the rest of its current task even after the person lets go. We formalize pHRI as a dynamical system, where the human has in mind an objective function they want the robot to optimize, but the robot does not get direct access to the parameters of this objective -- they are internal to the human. Within our proposed framework human interactions become observations about the true objective. We introduce approximations to learn from and respond to pHRI in real-time. We recognize that not all human corrections are perfect: often users interact with the robot noisily, and so we improve the efficiency of robot learning from pHRI by reducing unintended learning. Finally, we conduct simulations and user studies on a robotic manipulator to compare our proposed approach to the state-of-the-art. Our results indicate that learning from pHRI leads to better task performance and improved human satisfaction.

Soheil Habibian, Ananth Jonnavittula, Dylan P. Losey

Robots can learn from humans by asking questions. In these questions the robot demonstrates a few different behaviors and asks the human for their favorite. But how should robots choose which questions to ask? Today's robots optimize for informative questions that actively probe the human's preferences as efficiently as possible. But while informative questions make sense from the robot's perspective, human onlookers often find them arbitrary and misleading. In this paper we formalize active preference-based learning from the human's perspective. We hypothesize that -- from the human's point-of-view -- the robot's questions reveal what the robot has and has not learned. Our insight enables robots to use questions to make their learning process transparent to the human operator. We develop and test a model that robots can leverage to relate the questions they ask to the information these questions reveal. We then introduce a trade-off between informative and revealing questions that considers both human and robot perspectives: a robot that optimizes for this trade-off actively gathers information from the human while simultaneously keeping the human up to date with what it has learned. We evaluate our approach across simulations, online surveys, and in-person user studies. Videos of our user studies and results are available here: https://youtu.be/tC6y_jHN7Vw.

Siddharth Karamcheti, Albert J. Zhai, Dylan P. Losey et al.

It is challenging for humans -- particularly those living with physical disabilities -- to control high-dimensional, dexterous robots. Prior work explores learning embedding functions that map a human's low-dimensional inputs (e.g., via a joystick) to complex, high-dimensional robot actions for assistive teleoperation; however, a central problem is that there are many more high-dimensional actions than available low-dimensional inputs. To extract the correct action and maximally assist their human controller, robots must reason over their context: for example, pressing a joystick down when interacting with a coffee cup indicates a different action than when interacting with knife. In this work, we develop assistive robots that condition their latent embeddings on visual inputs. We explore a spectrum of visual encoders and show that incorporating object detectors pretrained on small amounts of cheap, easy-to-collect structured data enables i) accurately and robustly recognizing the current context and ii) generalizing control embeddings to new objects and tasks. In user studies with a high-dimensional physical robot arm, participants leverage this approach to perform new tasks with unseen objects. Our results indicate that structured visual representations improve few-shot performance and are subjectively preferred by users.

Mengxi Li, Alper Canberk, Dylan P. Losey et al.

When personal, assistive, and interactive robots make mistakes, humans naturally and intuitively correct those mistakes through physical interaction. In simple situations, one correction is sufficient to convey what the human wants. But when humans are working with multiple robots or the robot is performing an intricate task often the human must make several corrections to fix the robot's behavior. Prior research assumes each of these physical corrections are independent events, and learns from them one-at-a-time. However, this misses out on crucial information: each of these interactions are interconnected, and may only make sense if viewed together. Alternatively, other work reasons over the final trajectory produced by all of the human's corrections. But this method must wait until the end of the task to learn from corrections, as opposed to inferring from the corrections in an online fashion. In this paper we formalize an approach for learning from sequences of physical corrections during the current task. To do this we introduce an auxiliary reward that captures the human's trade-off between making corrections which improve the robot's immediate reward and long-term performance. We evaluate the resulting algorithm in remote and in-person human-robot experiments, and compare to both independent and final baselines. Our results indicate that users are best able to convey their objective when the robot reasons over their sequence of corrections.

Annie Xie, Dylan P. Losey, Ryan Tolsma et al.

Seamlessly interacting with humans or robots is hard because these agents are non-stationary. They update their policy in response to the ego agent's behavior, and the ego agent must anticipate these changes to co-adapt. Inspired by humans, we recognize that robots do not need to explicitly model every low-level action another agent will make; instead, we can capture the latent strategy of other agents through high-level representations. We propose a reinforcement learning-based framework for learning latent representations of an agent's policy, where the ego agent identifies the relationship between its behavior and the other agent's future strategy. The ego agent then leverages these latent dynamics to influence the other agent, purposely guiding them towards policies suitable for co-adaptation. Across several simulated domains and a real-world air hockey game, our approach outperforms the alternatives and learns to influence the other agent.

Ananth Jonnavittula, Dylan P. Losey

Assistive robots have the potential to help people perform everyday tasks. However, these robots first need to learn what it is their user wants them to do. Teaching assistive robots is hard for inexperienced users, elderly users, and users living with physical disabilities, since often these individuals are unable to show the robot their desired behavior. We know that inclusive learners should give human teachers credit for what they cannot demonstrate. But today's robots do the opposite: they assume every user is capable of providing any demonstration. As a result, these robots learn to mimic the demonstrated behavior, even when that behavior is not what the human really meant! Here we propose a different approach to reward learning: robots that reason about the user's demonstrations in the context of similar or simpler alternatives. Unlike prior works -- which err towards overestimating the human's capabilities -- here we err towards underestimating what the human can input (i.e., their choice set). Our theoretical analysis proves that underestimating the human's choice set is risk-averse, with better worst-case performance than overestimating. We formalize three properties to generate similar and simpler alternatives. Across simulations and a user study, our resulting algorithm better extrapolates the human's objective. See the user study here: https://youtu.be/RgbH2YULVRo

Mengxi Li, Dylan P. Losey, Jeannette Bohg et al.

When humans control drones, cars, and robots, we often have some preconceived notion of how our inputs should make the system behave. Existing approaches to teleoperation typically assume a one-size-fits-all approach, where the designers pre-define a mapping between human inputs and robot actions, and every user must adapt to this mapping over repeated interactions. Instead, we propose a personalized method for learning the human's preferred or preconceived mapping from a few robot queries. Given a robot controller, we identify an alignment model that transforms the human's inputs so that the controller's output matches their expectations. We make this approach data-efficient by recognizing that human mappings have strong priors: we expect the input space to be proportional, reversable, and consistent. Incorporating these priors ensures that the robot learns an intuitive mapping from few examples. We test our learning approach in robot manipulation tasks inspired by assistive settings, where each user has different personal preferences and physical capabilities for teleoperating the robot arm. Our simulated and experimental results suggest that learning the mapping between inputs and robot actions improves objective and subjective performance when compared to manually defined alignments or learned alignments without intuitive priors. The supplementary video showing these user studies can be found at: https://youtu.be/rKHka0_48-Q.

Erdem Bıyık, Dylan P. Losey, Malayandi Palan et al.

Reward functions are a common way to specify the objective of a robot. As designing reward functions can be extremely challenging, a more promising approach is to directly learn reward functions from human teachers. Importantly, data from human teachers can be collected either passively or actively in a variety of forms: passive data sources include demonstrations, (e.g., kinesthetic guidance), whereas preferences (e.g., comparative rankings) are actively elicited. Prior research has independently applied reward learning to these different data sources. However, there exist many domains where multiple sources are complementary and expressive. Motivated by this general problem, we present a framework to integrate multiple sources of information, which are either passively or actively collected from human users. In particular, we present an algorithm that first utilizes user demonstrations to initialize a belief about the reward function, and then actively probes the user with preference queries to zero-in on their true reward. This algorithm not only enables us combine multiple data sources, but it also informs the robot when it should leverage each type of information. Further, our approach accounts for the human's ability to provide data: yielding user-friendly preference queries which are also theoretically optimal. Our extensive simulated experiments and user studies on a Fetch mobile manipulator demonstrate the superiority and the usability of our integrated framework.

Hong Jun Jeon, Dylan P. Losey, Dorsa Sadigh

Assistive robots enable people with disabilities to conduct everyday tasks on their own. However, these tasks can be complex, containing both coarse reaching motions and fine-grained manipulation. For example, when eating, not only does one need to move to the correct food item, but they must also precisely manipulate the food in different ways (e.g., cutting, stabbing, scooping). Shared autonomy methods make robot teleoperation safer and more precise by arbitrating user inputs with robot controls. However, these works have focused mainly on the high-level task of reaching a goal from a discrete set, while largely ignoring manipulation of objects at that goal. Meanwhile, dimensionality reduction techniques for teleoperation map useful high-dimensional robot actions into an intuitive low-dimensional controller, but it is unclear if these methods can achieve the requisite precision for tasks like eating. Our insight is that---by combining intuitive embeddings from learned latent actions with robotic assistance from shared autonomy---we can enable precise assistive manipulation. In this work, we adopt learned latent actions for shared autonomy by proposing a new model structure that changes the meaning of the human's input based on the robot's confidence of the goal. We show convergence bounds on the robot's distance to the most likely goal, and develop a training procedure to learn a controller that is able to move between goals even in the presence of shared autonomy. We evaluate our method in simulations and an eating user study. See videos of our experiments here: https://youtu.be/7BouKojzVyk

Minae Kwon, Erdem Biyik, Aditi Talati et al.

In order to collaborate safely and efficiently, robots need to anticipate how their human partners will behave. Some of today's robots model humans as if they were also robots, and assume users are always optimal. Other robots account for human limitations, and relax this assumption so that the human is noisily rational. Both of these models make sense when the human receives deterministic rewards: i.e., gaining either $100 or $130 with certainty. But in real world scenarios, rewards are rarely deterministic. Instead, we must make choices subject to risk and uncertainty--and in these settings, humans exhibit a cognitive bias towards suboptimal behavior. For example, when deciding between gaining $100 with certainty or $130 only 80% of the time, people tend to make the risk-averse choice--even though it leads to a lower expected gain! In this paper, we adopt a well-known Risk-Aware human model from behavioral economics called Cumulative Prospect Theory and enable robots to leverage this model during human-robot interaction (HRI). In our user studies, we offer supporting evidence that the Risk-Aware model more accurately predicts suboptimal human behavior. We find that this increased modeling accuracy results in safer and more efficient human-robot collaboration. Overall, we extend existing rational human models so that collaborative robots can anticipate and plan around suboptimal human behavior during HRI.

Dylan P. Losey, Mengxi Li, Jeannette Bohg et al.

When teams of robots collaborate to complete a task, communication is often necessary. Like humans, robot teammates should implicitly communicate through their actions: but interpreting our partner's actions is typically difficult, since a given action may have many different underlying reasons. Here we propose an alternate approach: instead of not being able to infer whether an action is due to exploration, exploitation, or communication, we define separate roles for each agent. Because each role defines a distinct reason for acting (e.g., only exploit, only communicate), teammates now correctly interpret the meaning behind their partner's actions. Our results suggest that leveraging and alternating roles leads to performance comparable to teams that explicitly exchange messages. You can find more images and videos of our experimental setups at http://ai.stanford.edu/blog/learning-from-partners/.

Erdem Bıyık, Malayandi Palan, Nicholas C. Landolfi et al.

Robots can learn the right reward function by querying a human expert. Existing approaches attempt to choose questions where the robot is most uncertain about the human's response; however, they do not consider how easy it will be for the human to answer! In this paper we explore an information gain formulation for optimally selecting questions that naturally account for the human's ability to answer. Our approach identifies questions that optimize the trade-off between robot and human uncertainty, and determines when these questions become redundant or costly. Simulations and a user study show our method not only produces easy questions, but also ultimately results in faster reward learning.

Dylan P. Losey, Krishnan Srinivasan, Ajay Mandlekar et al.

Assistive robotic arms enable users with physical disabilities to perform everyday tasks without relying on a caregiver. Unfortunately, the very dexterity that makes these arms useful also makes them challenging to teleoperate: the robot has more degrees-of-freedom than the human can directly coordinate with a handheld joystick. Our insight is that we can make assistive robots easier for humans to control by leveraging latent actions. Latent actions provide a low-dimensional embedding of high-dimensional robot behavior: for example, one latent dimension might guide the assistive arm along a pouring motion. In this paper, we design a teleoperation algorithm for assistive robots that learns latent actions from task demonstrations. We formulate the controllability, consistency, and scaling properties that user-friendly latent actions should have, and evaluate how different low-dimensional embeddings capture these properties. Finally, we conduct two user studies on a robotic arm to compare our latent action approach to both state-of-the-art shared autonomy baselines and a teleoperation strategy currently used by assistive arms. Participants completed assistive eating and cooking tasks more efficiently when leveraging our latent actions, and also subjectively reported that latent actions made the task easier to perform. The video accompanying this paper can be found at: https://youtu.be/wjnhrzugBj4.

Dylan P. Losey, Dorsa Sadigh

Humans often assume that robots are rational. We believe robots take optimal actions given their objective; hence, when we are uncertain about what the robot's objective is, we interpret the robot's actions as optimal with respect to our estimate of its objective. This approach makes sense when robots straightforwardly optimize their objective, and enables humans to learn what the robot is trying to achieve. However, our insight is that---when robots are aware that humans learn by trusting that the robot actions are rational---intelligent robots do not act as the human expects; instead, they take advantage of the human's trust, and exploit this trust to more efficiently optimize their own objective. In this paper, we formally model instances of human-robot interaction (HRI) where the human does not know the robot's objective using a two-player game. We formulate different ways in which the robot can model the uncertain human, and compare solutions of this game when the robot has conservative, optimistic, rational, and trusting human models. In an offline linear-quadratic case study and a real-time user study, we show that trusting human models can naturally lead to communicative robot behavior, which influences end-users and increases their involvement.

Dylan P. Losey, Marcia K. O'Malley

During human-robot interaction (HRI), we want the robot to understand us, and we want to intuitively understand the robot. In order to communicate with and understand the robot, we can leverage interactions, where the human and robot observe each other's behavior. However, it is not always clear how the human and robot should interpret these actions: a given interaction might mean several different things. Within today's state-of-the-art, the robot assigns a single interaction strategy to the human, and learns from or teaches the human according to this fixed strategy. Instead, we here recognize that different users interact in different ways, and so one size does not fit all. Therefore, we argue that the robot should maintain a distribution over the possible human interaction strategies, and then infer how each individual end-user interacts during the task. We formally define learning and teaching when the robot is uncertain about the human's interaction strategy, and derive solutions to both problems using Bayesian inference. In examples and a benchmark simulation, we show that our personalized approach outperforms standard methods that maintain a fixed interaction strategy.

Dylan P. Losey, Marcia K. O'Malley

It is difficult for humans to efficiently teach robots how to correctly perform a task. One intuitive solution is for the robot to iteratively learn the human's preferences from corrections, where the human improves the robot's current behavior at each iteration. When learning from corrections, we argue that while the robot should estimate the most likely human preferences, it should also know what it does not know, and integrate this uncertainty as it makes decisions. We advance the state-of-the-art by introducing a Kalman filter for learning from corrections: this approach obtains the uncertainty of the estimated human preferences. Next, we demonstrate how the estimate uncertainty can be leveraged for active learning and risk-sensitive deployment. Our results indicate that obtaining and leveraging uncertainty leads to faster learning from human corrections.

Dylan P. Losey, Marcia K. O'Malley

Robots are finding new applications where physical interaction with a human is necessary: manufacturing, healthcare, and social tasks. Accordingly, the field of physical human-robot interaction (pHRI) has leveraged impedance control approaches, which support compliant interactions between human and robot. However, a limitation of traditional impedance control is that---despite provisions for the human to modify the robot's current trajectory---the human cannot affect the robot's future desired trajectory through pHRI. In this paper, we present an algorithm for physically interactive trajectory deformations which, when combined with impedance control, allows the human to modulate both the actual and desired trajectories of the robot. Unlike related works, our method explicitly deforms the future desired trajectory based on forces applied during pHRI, but does not require constant human guidance. We present our approach and verify that this method is compatible with traditional impedance control. Next, we use constrained optimization to derive the deformation shape. Finally, we describe an algorithm for real time implementation, and perform simulations to test the arbitration parameters. Experimental results demonstrate reduction in the human's effort and improvement in the movement quality when compared to pHRI with impedance control alone.