Learning-based motion planning can quickly generate near-optimal trajectories. However, it often requires either large training datasets or costly collection of human demonstrations. This work proposes an alternative approach that quickly generates smooth, near-optimal collision-free 3D Cartesian trajectories from a single artificial demonstration. The demonstration is encoded as a Dynamic Movement Primitive (DMP) and iteratively reshaped using policy-based reinforcement learning guided by model-specific cost designs, resulting in a diverse trajectory dataset for varying obstacle configurations. This dataset is used to train a neural network that takes as inputs the task parameters describing the obstacle dimensions and location, derived automatically from a point cloud, and outputs the DMP parameters that generate the trajectory. The approach is validated in simulation and real-robot experiments, outperforming a RRT-Connect baseline and achieving comparable performance to the CHOMP planner, while supporting multi-modal trajectory generation for different obstacle geometries and end-effector dimensions.

Symbolic representation learning from unsupervised environment interactions is a promising strategy for enabling autonomous agents to solve long-horizon decision problems. These approaches popularly use encoder-decoder architectures with Gumbel-reparameterized categorical latents to discover predicates for use with classical planners. However, if the number of symbols to be learned is unknown, the emergence of overly fine-grained or redundant symbols can lead to an overabundance of distinct propositional states, which causes the number of PDDL actions in the generated planning domains to increase substantially. To mitigate this issue, we propose CompactSym, a regularization procedure that employs constrained minimization of the estimated marginal entropy of the discrete latent variables. This enables learning significantly more compact symbolic representations and planning domains while still preserving the task-relevant information by bounding decoding loss. In experiments on a visual gridwalk task and a robotic multi-object manipulation environment, CompactSym consistently decreases the number of discrete latent states used, and speeds up planning without any sacrifices in task success rate. Our code is available at https://github.com/henrijacgeiss/compact_sym/ .

Traditional task-and-motion planning (TAMP) approaches primarily focus on defining sequences of actions along with the necessary geometric and kinematic constraints to execute long-horizon tasks. However, their applicability in real-world settings is limited, as they typically assume simplified object models that overlook key physical properties critical for the successful execution of contact-rich tasks. Moreover, they often use sub-symbolic reasoning during motion planning, which drastically increases planning time and decreases overall success rates. We propose a method that leverages a TAMP approach, defining object-centric abstractions of execution constraints, called Unified TAMP (U-TAMP), to execute robotic tasks involving interactions among objects with heterogeneous shapes, sizes, and materials. Using a Vision-Language Model (VLM), we generate abstractions of inter-object affordances for characterizing physical interaction constraints between objects in contact-rich tasks, such as grasp and support constraints. These constraints are used to enrich the U-TAMP planning domain to deal with objects with variable physical properties. We perform experiments in simulated kitchen table organization scenarios and compare our results with those of the original U-TAMP, as well as a state-of-the-art VLM-based planner that leverages common sense knowledge of objects' affordances for plan generation. Our approach achieves significantly higher planning success rates and improves planning times by one to two orders of magnitude compared to other methods.

We introduce a new method that extracts knowledge from a large language model (LLM) to produce object-level plans, which describe high-level changes to object state, and uses them to bootstrap task and motion planning (TAMP). Existing work uses LLMs to directly output task plans or generate goals in representations like PDDL. However, these methods fall short because they rely on the LLM to do the actual planning or output a hard-to-satisfy goal. Our approach instead extracts knowledge from an LLM in the form of plan schemas as an object-level representation called functional object-oriented networks (FOON), from which we automatically generate PDDL subgoals. Our method markedly outperforms alternative planning strategies in completing several pick-and-place tasks in simulation.(dagger)

Following work on joint object-action representations, functional object-oriented networks (FOON) were introduced as a knowledge graph representation for robots. A FOON contains symbolic concepts useful to a robot's understanding of tasks and its environment for object-level planning. Prior to this work, little has been done to show how plans acquired from FOON can be executed by a robot, as the concepts in a FOON are too abstract for execution. We thereby introduce the idea of exploiting object-level knowledge as a FOON for task planning and execution. Our approach automatically transforms FOON into PDDL and leverages off-the-shelf planners, action contexts, and robot skills in a hierarchical planning pipeline to generate executable task plans. We demonstrate our entire approach on long-horizon tasks in CoppeliaSim and show how learned action contexts can be extended to never-before-seen scenarios.

In task and motion planning (TAMP), the ambiguity and underdetermination of abstract descriptions used by task planning methods make it difficult to characterize physical constraints needed to successfully execute a task. The usual approach is to overlook such constraints at task planning level and to implement expensive sub-symbolic geometric reasoning techniques that perform multiple calls on unfeasible actions, plan corrections, and re-planning until a feasible solution is found. We propose an alternative TAMP approach that unifies task and motion planning into a single heuristic search. Our approach is based on an object-centric abstraction of motion constraints that permits leveraging the computational efficiency of off-the-shelf AI heuristic search to yield physically feasible plans. These plans can be directly transformed into object and motion parameters for task execution without the need of intensive sub-symbolic geometric reasoning.

—Following work on joint object-action representations, functional object-oriented networks (FOON) were introduced as a knowledge graph representation for robots. Taking the form of a bipartite graph, a FOON contains symbolic (high-level) concepts useful to a robot’s understanding of tasks and its environment for object-level planning. Prior to this paper, little has been done to demonstrate how task plans acquired from FOON via task tree retrieval can be executed by a robot, as the concepts in a FOON are too abstract for immediate execution. We propose a hierarchical task planning approach that translates a FOON graph into a PDDL-based representation of domain knowledge for manipulation planning. As a result of this process, a task plan can be acquired that a robot can execute from start to end, leveraging the use of action contexts and skills in the form of dynamic movement primitives (DMP). We demonstrate the entire pipeline from planning to execution using CoppeliaSim and show how learned action contexts can be extended to never-before-seen scenarios.

—Following work on joint object-action representa- tion, functional object-oriented networks (FOON) were introduced as a knowledge representation for robots. A FOON contains symbolic (high-level) concepts useful to a robot’s understanding of tasks and its environment for object-level planning. Prior to this work, little has been done to show how plans acquired from FOON can be executed by a robot, as the concepts in a FOON are too abstract for immediate execution. We propose a hierarchical task planning approach that translates a FOON graph into a PDDL-based representation of domain knowledge for manipulation planning. As a result of this process, a manipulation plan can be acquired, which can be executed by a robot from start to end, leveraging the use of action contexts and skills as dynamic movement primitives (DMPs). We demonstrate the entire pipeline from planning to execution using CoppeliaSim and show how learned action contexts can be extended to never-before-seen scenarios.

The success of deep reinforcement learning approaches to learn dexterous manipulation skills strongly hinges on the rewards assigned to actions during task execution. The usual approach is to handcraft the reward function but due to the high complexity of dexterous manipulations the reward definition demands large engineering effort for each particular task. To avoid this burden, we use an inverse reinforcement learning (IRL) approach to automatically learn the reward function using samples obtained from demonstrations of desired behaviours. We have identified that the learned rewards using existing IRL approaches are strongly biased towards demonstrated actions due to the scarcity of samples in the vast state-action space of dexterous manipulation applications. This significantly hinders performance due to unreliable reward estimations in regions unexplored during demonstration. We use statistical tools for random sample generation and reward normalization to reduce this bias. We show that this approach improves learning stability and transferability of IRL for dexterous manipulation tasks. Project page: https://sites.google.con view/irl-for-dexterous-hand

Task and motion planning deals with complex tasks that require a robot to automatically define and execute multi-step sequences of actions in cluttered scenarios. In this context, a linear motion is often not sufficient to approach a target object since collisions of the gripper with other objects or the target object might occur. Thus, motion planners should be able to generate collision-free trajectories for every particular configuration of obstacles for grounding the symbolic actions of the task plan. Current approaches either search for feasible motions offline using computationally expensive trial-and-error processes on physically realistic simulations or learn a set of motion parameters for particular object configuration spaces with little generalization. This work proposes an appealing alternative by efficiently generating trajectories for the collision-free execution of symbolic actions in variable scenarios without the need of intensive offline simulations. Our approach combines the benefit of learning from demonstration, to quickly generate an initial set of motion parameters for each symbolic action, with policy improvement with path integrals, to diversify this initial set of parameters to cope with different obstacle configurations. We show how the improved flexibility is achieved after a few minutes of training and successfully solves tasks requiring different sequences of picking and placing actions in variable configurations of obstacles.

Following work on joint object-action representations, the functional object-oriented network (FOON) was introduced as a knowledge graph representation for robots. Taking the form of a bipartite graph, a FOON contains symbolic or high-level information that would be pertinent to a robot's understanding of its environment and tasks in a way that mirrors human understanding of actions. In this work, we outline a road-map for future development of FOON and its application in robotic systems for task planning as well as knowledge acquisition from demonstration. We propose preliminary ideas to show how a FOON can be created in a real-world scenario with a robot and human teacher in a way that can jointly augment existing knowledge in a FOON and teach a robot the skills it needs to replicate the demonstrated actions and solve a given manipulation problem.

Current approaches combining task and motion planning require intensive geometric and symbolic reasoning to find feasible motions for task execution. The poor expressiveness of task planning domains for characterizing geometric changes with actions and the difficulties faced by current approaches to efficiently identify motion dependencies for plan execution produce expensive callings to motion planning on unfeasible actions and intensive reasoning to find realizable plans. In this work we combine two recent approaches to address these problems. Task planning is carried out using an object-centered description of geometric relations that consistently characterizes changes in the object configuration space. Plan execution is implemented using a symbol to motion hierarchical decomposition that depends on consecutive actions in the plan, rather than on single actions, which permits considering motion dependencies across plan actions for a successful execution.

The definition of symbolic descriptions that consistently represent relevant geometrical aspects in manipulation tasks is a challenging problem that has received little attention in the robotic community. This definition is usually done from an observer perspective of a finite set of object relations and orientations that only satisfy geometrical constraints to execute experiments in laboratory conditions. This restricts the possible changes with manipulation actions in the object configuration space to those compatible with that particular external reference definitions, which greatly limits the spectrum of possible manipulations. To tackle these limitations we propose an object-centered representation that permits characterizing a much wider set of possible changes in configuration spaces than the traditional observer perspective counterpart. Based on this representation, we define universal planning operators for picking and placing actions that permits generating plans with geometric and force consistency in manipulation tasks. This object-centered description is directly obtained from the poses and bounding boxes of objects using a novel learning mechanisms that permits generating signal-symbols relations without the need of handcrafting these relations for each particular scenario.

The distributional hypothesis states that the meaning of a concept is defined through the contexts it occurs in. In practice, often word co-occurrence and proximity are analyzed in text corpora for a given word to obtain a real-valued semantic word vector, which is taken to (at least partially) encode the meaning of this word. Here we transfer this idea from text to images, where pre-assigned labels of other objects or activations of convolutional neural networks serve as context. We propose a simple algorithm that extracts and processes object contexts from an image database and yields semantic vectors for objects. We show empirically that these representations exhibit on par performance with state-of-the-art distributional models over a set of conventional objects. For this we employ well-known word benchmarks in addition to a newly proposed object-centric benchmark.

Oil and coconut palm trees are important crops in many tropical countries, which are either planted as plantations or scattered in the landscape. Monitoring in terms of counting provides useful information for various stakeholders. Most of the existing monitoring methods are based on spectral profiles or simple neural networks and either fall short in terms of accuracy or speed. We use a neural network of the U-Net type in order to detect oil and coconut palms on very high resolution satellite images. The method is applied to two different study areas: (1) large monoculture oil palm plantations in Jambi, Indonesia, and (2) coconut palms in the Bengaluru Metropolitan Region in India. The results show that the proposed method reaches a performance comparable to state of the art approaches, while being about one order of magnitude faster. We reach a maximum throughput of 235 ha/s with a spatial image resolution of 40 cm. The proposed method proves to be reliable even under difficult conditions, such as shadows or urban areas, and can easily be transferred from one region to another. The method detected palms with accuracies between 89% and 92%.

Function approximation in online, incremental, reinforcement learning needs to deal with two fundamental problems: biased sampling and nonstationarity. In this kind of task, biased sampling occurs because samples are obtained from specific trajectories dictated by the dynamics of the environment and are usually concentrated in particular convergence regions, which in the long term tend to dominate the approximation in the less sampled regions. The nonstationarity comes from the recursive nature of the estimations typical of temporal difference methods. This nonstationarity has a local profile, varying not only along the learning process but also along different regions of the state space. We propose to deal with these problems using an estimation of the probability density of samples represented with a gaussian mixture model. To deal with the nonstationarity problem, we use the common approach of introducing a forgetting factor in the updating formula. However, instead of using the same forgetting factor for the whole domain, we make it dependent on the local density of samples, which we use to estimate the nonstationarity of the function at any given input point. To address the biased sampling problem, the forgetting factor applied to each mixture component is modulated according to the new information provided in the updating, rather than forgetting depending only on time, thus avoiding undesired distortions of the approximation in less sampled regions.

In large industrial greenhouses, plants are usually treated following well established protocols for watering, nutrients, and shading/light. While this is practical for the automation of the process, it does not tap the full potential for optimal plant treatment. To more efficiently grow plants, specific treatments according to the plant individual needs should be applied. Experienced human gardeners are very good at treating plants individually. Unfortunately, hiring a crew of gardeners to carry out this task in large greenhouses is not cost effective. In this work we present a cognitive system that integrates artificial intelligence (AI) techniques for decision-making with robotics techniques for sensing and acting to autonomously treat plants using a real-robot platform. Artificial intelligence techniques are used to decide the amount of water and nutrients each plant needs according to the history of the plant. Robotic techniques for sensing measure plant attributes (e.g. leaves) from visual information using 3D model representations. These attributes are used by the AI system to make decisions about the treatment to apply. Acting techniques execute robot movements to supply the plants with the specified amount of water and nutrients.

In this paper, we address the problem of finding replacements of missing objects, involved in the execution of manipulation tasks. Our approach is based on estimating functional affordances for the unknown objects in order to propose replacements. We use a vision-based affordance estimation system utilizing object-wise global features and a multi-label learning method. This method also associates confidence values to the estimated affordances. We evaluate our approach on kitchen-related manipulation affordances. The evaluation also includes testing different scenarios for training the system using large-scale datasets. The results indicate that the system is able to successfully predict the affordances of novel objects. We also implement our system on a humanoid robot and demonstrate the affordance estimation in a real scene.

In the online version of the EM algorithm introduced by Sato and Ishii ( 2000 ), a time-dependent discount factor is introduced for forgetting the effect of the old estimated values obtained with an earlier, inaccurate estimator. In their approach, forgetting is uniformly applied to the estimators of each mixture component depending exclusively on time, irrespective of the weight attributed to each unit for the observed sample. This causes an excessive forgetting in the less frequently sampled regions. To address this problem, we propose a modification of the algorithm that involves a weight-dependent forgetting, different for each mixture component, in which old observations are forgotten according to the actual weight of the new samples used to replace older values. A comparison of the time-dependent versus the weight-dependent approach shows that the latter improves the accuracy of the approximation and exhibits much greater stability.