Multi-task optimization is a powerful approach for solving a large number of tasks in parallel. However, existing algorithms face distinct limitations: Population-based methods scale poorly and remain underexplored for large task sets. Approaches that do scale beyond a thousand tasks are mostly MAP-Elites variants and rely on a fixed, discretized archive that disregards the topology of the task space. We introduce MONET (Multi-Task Optimization over Networks of Tasks), a multi-task optimization algorithm that models the task space as a graph: tasks are nodes, and edges connect tasks in the task parameter space. This representation enables knowledge transfer between tasks and remains tractable for high-dimensional problems while exploiting the topology of the task space. MONET combines social learning, which generates candidates from neighboring nodes via crossover, with individual learning, which refines a node’s own solution independently via mutation. We evaluate MONET on four domains (archery, arm, and cartpole with 5,000 tasks each; hexapod with 2,000 tasks) and show that it matches or exceeds the performance of existing MAP–Elites-based baselines across all four domains.

Evolution and learning have historically been interrelated topics, and their interplay is attracting increased interest lately. The emerging new factor in this trend is morphological evolution, the evolution of physical forms within embodied AI systems such as robots. In this study, we investigate a system of hexacopter-type drones with evolvable morphologies and learnable controllers and make contributions to two fields. For aerial robotics, we demonstrate that the combination of evolution and learning can deliver non-conventional drones that significantly outperform the traditional hexacopter on several tasks that are more complex than previously considered in the literature. For the field of Evolutionary Computing, we introduce novel metrics and perform new analyses into the interaction of morphological evolution and learning, uncovering hitherto unidentified effects. Our analysis tools are domain-agnostic, making a methodological contribution towards building solid foundations for embodied AI systems that integrate evolution and learning.

This study explores the integration of Lamarckian system into evolutionary robotics (ER), comparing it with the traditional Darwinian model across various environments. By adopting Lamarckian principles, where robots inherit learned traits, alongside Darwinian learning without inheritance, we investigate adaptation in dynamic settings. Our research, conducted in six distinct environmental setups, demonstrates that Lamarckian systems outperform Darwinian ones in adaptability and efficiency, particularly in challenging conditions. Our analysis highlights the critical role of the interplay between controller \& morphological evolution and environment adaptation, with parent-offspring similarities and newborn \&survivors before and after learning providing insights into the effectiveness of trait inheritance. Our findings suggest Lamarckian principles could significantly advance autonomous system design, highlighting the potential for more adaptable and robust robotic solutions in complex, real-world applications. These theoretical insights were validated using real physical robots, bridging the gap between simulation and practical application.

This paper studies the effects of extending an evolvable modular robot system by 'bones' of evolvable length. Based on simulations, we investigate whether evolution exploits this possibility, and we compare different ways to attain desired morphological features alongside preferred behavioral traits. To validate the outcomes, we select a robot with a high fitness and interesting morphology and construct its physical twin. Comparing the simulated and real-world behaviors of the twins, we gain insight into the reality gap.

Learning from Demonstrations (LfD) allows robots to learn skills from human users, but its effectiveness can suffer due to sub-optimal teaching, especially from untrained demonstrators. Active LfD aims to improve this by letting robots actively request demonstrations to enhance learning. However, this may lead to frequent context switches between various task situations, increasing the human cognitive load and introducing errors to demonstrations. Moreover, few prior studies in active LfD have examined how these active query strategies may impact human teaching in aspects beyond user experience, which can be crucial for developing algorithms that benefit both robot learning and human teaching. To tackle these challenges, we propose an active LfD method that optimizes the query sequence of online human demonstrations via Curriculum Learning (CL), where demonstrators are guided to provide demonstrations in situations of gradually increasing difficulty. We evaluate our method across four simulated robotic tasks with sparse rewards and conduct a user study (N=26) to investigate the influence of active LfD methods on human teaching regarding teaching performance, post-guidance teaching adaptivity, and teaching transferability. Our results show that our method significantly improves learning performance compared to three other LfD baselines in terms of the final success rate of the converged policy and sample efficiency. Additionally, results from our user study indicate that our method significantly reduces the time required from human demonstrators and decreases failed demonstration attempts. It also enhances post-guidance human teaching in both seen and unseen scenarios compared to another active LfD baseline, indicating enhanced teaching performance, greater post-guidance teaching adaptivity, and better teaching transferability achieved by our method.

In this paper, we compare Bayesian Optimization, Differential Evolution, and an Evolution Strategy employed as a gait-learning algorithm in modular robots. The motivational scenario is the joint evolution of morphologies and controllers, where "newborn" robots also undergo a learning process to optimize their inherited controllers (without changing their bodies). This context raises the question: How do gait-learning algorithms compare when applied to various morphologies that are not known in advance (and thus need to be treated as without priors)? To answer this question, we use a test suite of twenty different robot morphologies to evaluate our gait-learners and compare their efficiency, efficacy, and sensitivity to morphological differences. The results indicate that Bayesian Optimization and Differential Evolution deliver the same solution quality (walking speed for the robot) with fewer evaluations than the Evolution Strategy. Furthermore, the Evolution Strategy is more sensitive for morphological differences (its efficacy varies more between different morphologies) and is more subject to luck (repeated runs on the same morphology show greater variance in the outcomes).

When controllers (brains) and morphologies (bodies) of robots simultaneously evolve, this can lead to a problem, namely the brain body mismatch problem. In this research, we propose a solution of lifetime learning. We set up a system where modular robots can create offspring that inherit the bodies of parents by recombination and mutation. With regards to the brains of the offspring, we use two methods to create them. The first one entails solely evolution which means the brain of a robot child is inherited from its parents. The second approach is evolution plus learning which means the brain of a child is inherited as well, but additionally is developed by a learning algorithm - RevDEknn. We compare these two methods by running experiments in a simulator called Revolve and use efficiency, efficacy, and the morphology intelligence of the robots for the comparison. The experiments show that the evolution plus learning method does not only lead to a higher fitness level, but also to more morphologically evolving robots. This constitutes a quantitative demonstration that changes in the brain can induce changes in the body, leading to the concept of morphological intelligence, which is quantified by the learning delta, meaning the ability of a morphology to facilitate learning.

The discrepancy between simulated and hardware experiments, the reality gap, is a challenge in evolutionary robotics. While strategies have been proposed to address this gap in fixed-body robots, they are not viable when dealing with populations and generations where the body is in constant change. The continual evolution of body designs necessitates the manufacturing of new robotic structures, a process that can be time-consuming if carried out manually. Moreover, the increased manufacturing time not only prolongs hardware experimental durations but also disrupts the synergy between hardware and simulated experiments. Failure to effectively manage these challenges could impede the implementation of evolutionary robotics in real-life environments. The Autonomous Robot Evolution project presents a framework to tackle these challenges through a case study. This paper describes the main three contributions of this work: Firstly, it analyses the different reality gap experienced by each different robot or the heterogenous reality gap. Secondly, it emphasizes the importance of automation in robot manufacturing. And thirdly, it highlights the necessity of a framework to orchestrate the synergy between simulated and hardware experiments. In the long term, integrating these contributions into evolutionary robotics is envisioned to enable the continuous production of robots in real-world environments.

Natural groups of animals, such as swarms of social insects, exhibit astonishing degrees of task specialization, useful to address complex tasks and to survive. This is supported by phenotypic plasticity: individuals sharing the same genotype that is expressed differently for different classes of individuals, each specializing in one task. In this work, we evolve a swarm of simulated robots with phenotypic plasticity to study the emergence of specialized collective behavior during an emergent perception task. Phenotypic plasticity is realized in the form of heterogeneity of behavior by dividing the genotype into two components, with one different neural network controller associated to each component. The whole genotype, expressing the behavior of the whole group through the two components, is subject to evolution with a single fitness function. We analyse the obtained behaviors and use the insights provided by these results to design an online regulatory mechanism. Our experiments show three main findings: 1) The sub-groups evolve distinct emergent behaviors. 2) The effectiveness of the whole swarm depends on the interaction between the two sub-groups, leading to a more robust performance than with singular sub-group behavior. 3) The online regulatory mechanism enhances overall performance and scalability.

We investigate an evolvable robot system where the body provides proprioceptive sensory signals to the controller (brain) about the positions of the joints. The key aspect we consider is whether all joints should be sensed or if sensing fewer joints would be better. We research this matter based on a test suite of twenty-two robots with various shapes and sizes and implement a system where the controller and the sensory signal system evolve together. Experiments with this system show that the evolved solutions use signals only from a fraction of the joints (25–51

Legged robots are well-suited for deployment in unstructured environments but require a unique control scheme specific for their design. As controllers optimised in simulation do not transfer well to the real world (the infamous sim-to-real gap), methods enabling quick learning in the real world, without any assumptions on the specific robot model and its dynamics, are necessary. In this paper, we present a generic method based on Central Pattern Generators, that enables the acquisition of basic locomotion skills in parallel, through very few trials. The novelty of our approach, underpinned by a mathematical analysis of the controller model, is to search for good initial states, instead of optimising connection weights. Empirical validation in six different robot morphologies demonstrates that our method enables robots to learn primary locomotion skills in less than 15 minutes in the real world. In the end, we showcase our skills in a targeted locomotion experiment. The authors present a machine learning approach to enable modular robots to learn to walk and follow objects within a short time. The method is based on bioinspired central pattern generators and is validated on six robots with different body shapes.

In this paper, we present a new method for a swarm to collectively sense and follow a gradient in the environment. The agents in the swarm only rely on relative distance and bearing measurements of neighbors. Additionally, only a minority of agents in the swarm perceive the scalar value of the gradient at their location. We test the method with incrementally changing ratio of agents with sensors on various swarm sizes. In addition to repeated simulation experiments, we also test the performance with a real nano-drone swarm. Results show us that, using the new method, the swarm was successful at following the gradient in the environment even with a low portion of the swarm with sensors on various swarm sizes. A real nano-drone swarm also demonstrates a good performance in our test even with members having disabled sensors.

Reinforcement Learning (RL) has achieved great success in sequential decision-making problems but often requires extensive agent-environment interactions. To improve sample efficiency, methods like Reinforcement Learning from Expert Demonstrations (RLED) incorporate external expert demonstrations to aid agent exploration during the learning process. However, these demonstrations, typically collected from human users, are costly and thus often limited in quantity. Therefore, how to select the optimal set of human demonstrations that most effectively aids learning becomes a critical concern. This paper introduces EARLY (Episodic Active Learning from demonstration querY), an algorithm designed to enable a learning agent to generate optimized queries for expert demonstrations in a trajectory-based feature space. EARLY employs a trajectory-level estimate of uncertainty in the agent’s current policy to determine the optimal timing and content for feature-based queries. By querying episodic demonstrations instead of isolated state-action pairs, EARLY enhances the human teaching experience and achieves better learning performance. We validate the effectiveness of our method across three simulated navigation tasks of increasing difficulty. Results indicate that our method achieves expert-level performance in all three tasks, converging over 50% faster than other four baseline methods when demonstrations are generated by simulated oracle policies. A follow-up pilot user study (N = 18) further supports that our method maintains significantly better convergence with human expert demonstrators, while also providing a better user experience in terms of perceived task load and requiring significantly less human time.

Jointly optimising both the body and brain of a robot is known to be a challenging task, especially when attempting to evolve designs in simulation that will subsequently be built in the real world. To address this, it is increasingly common to combine evolution with a learning algorithm that can either improve the inherited controllers of new offspring to fine tune them to the new body design or learn them from scratch. In this paper an approach is proposed in which a robot is specified indirectly by two compositional pattern producing networks (CPPN) encoded in a single genome, one which encodes the brain and the other the body. The body part of the genome is evolved using an evolutionary algorithm (EA), with an individual learning algorithm (also an EA) applied to the inherited controller to improve it. The goal of this paper is to determine how to utilise the results of learning process most effectively to improve task performance of the robot. Specifically, three variants are investigated: (1) evolution of the body+controller only; (2) a learning algorithm is applied to the inherited controller with the learned fitness assigned to the genome; (3) learning is applied and the genome is updated with the learned controller, as well as being assigned the learned fitness. Experiments are performed in three different scenarios chosen to favour different bodies and locomotion patterns. It is shown that better performance can be obtained using learning but only if the learned controller is inherited by the offspring.

This paper investigates human–robot collaboration in a novel setup: a human helps a mobile robot that can move and navigate freely in an environment. Specifically, the human helps by remotely taking over control during the learning of a task. The task is to find and collect several items in a walled arena, and Reinforcement Learning is used to seek a suitable controller. If the human observes undesired robot behavior, they can directly issue commands for the wheels through a game joystick. Experiments in a simulator showed that human assistance improved robot behavior efficacy by 30% and efficiency by 12%. The best policies were also tested in real life, using physical robots. Hardware experiments showed no significant difference concerning the simulations, providing empirical validation of our approach in practice.

The main question this paper addresses is: What combination of a robot controller and a learning method should be used, if the morphology of the learning robot is not known in advance? Our interest is rooted in the context of morphologically evolving modular robots, but the question is also relevant in general, for system designers interested in widely applicable solutions. We perform an experimental comparison of three controller-and-learner combinations: one approach where controllers are based on modelling animal locomotion (Central Pattern Generators, CPG) and the learner is an evolutionary algorithm, a completely different method using Reinforcement Learning (RL) with a neural network controller architecture, and a combination `in-between' where controllers are neural networks and the learner is an evolutionary algorithm. We apply these three combinations to a test suite of modular robots and compare their efficacy, efficiency, and robustness. Surprisingly, the usual CPG-based and RL-based options are outperformed by the in-between combination that is more robust and efficient than the other two setups.

Evolutionary robot systems offer two principal advantages: an advanced way of developing robots through evolutionary optimization and a special research platform to conduct what-if experiments regarding questions about evolution. Our study sits at the intersection of these. We investigate the question "What if the 18th-century biologist Lamarck was not completely wrong and individual traits learned during a lifetime could be passed on to offspring through inheritance?" We research this issue through simulations with an evolutionary robot framework where morphologies (bodies) and controllers (brains) of robots are evolvable and robots also can improve their controllers through learning during their lifetime. Within this framework, we compare a Lamarckian system, where learned bits of the brain are inheritable, with a Darwinian system, where they are not. Analyzing simulations based on these systems, we obtain new insights about Lamarckian evolution dynamics and the interaction between evolution and learning. Specifically, we show that Lamarckism amplifies the emergence of 'morphological intelligence', the ability of a given robot body to acquire a good brain by learning, and identify the source of this success: newborn robots have a higher fitness because their inherited brains match their bodies better than those in a Darwinian system.

In Evolutionary Robotics where both body and brain are malleable, it is common practice to evaluate individuals in isolated environments. With the objective of implementing a more naturally plausible system, we designed a single interactive ecosystem for robots to be evaluated in. In this ecosystem robots are physically present and can interact each other and we implemented decentralized rules for mate selection and reproduction. To study the effects of evaluating robots in an interactive ecosystem has on evolution, we compare the evolutionary process with a more traditional, oracle–based approach. In our analysis, we observe how the different approach has a substantial impact on the final behaviour and morphology of the robots, while maintaining decent fitness performance.

Human-centered interactive robot tasks (e.g., social greetings and cooperative dressing) are a type of task where humans are involved in task dynamics and performance evaluation. Such tasks require spatial and temporal coordination between agents in real-time, tackling physical limitations from constrained robot bodies, and connecting human user experience with concrete learning objectives to inform algorithm design. To solve these challenges, imitation learning has become a popular approach where by a robot learns to perform a task by imitating how human experts do it (i.e., expert policies). However, previous works tend to isolate the algorithm design from the design of the whole learning pipeline, neglecting its connection with other modules inside the process (like data collection and user-centered subjective evaluation) from the view as a system. Going beyond traditional imitation learning, this work reexamines robot imitation learning in human-centered interactive tasks from the perspective of the whole learning pipeline, ranging from data collection to subjective evaluation. We present a process-oriented framework that consists of a guideline to collect diverse yet representative demonstrations and an interpreter to explain subjective user-centered performance with objective robot-related parameters. We illustrate the steps covered by the framework in a fist-bump greeting task as demonstrative deployment. Results show that our framework is able to identify representative human-centered features to instruct demonstration collection and validate influential robot-centered factors to interpret the gap in subjective performance between the expert policy and the imitator policy.