Deep learning has revolutionised many fields, but it is still challenging to transfer its success to small mobile robots with minimal hardware. Specifically, some work has been done to this effect in the RoboCup humanoid football domain, but results that are performant and efficient and still generally applicable outside of this domain are lacking. We propose an approach conceptually different from those taken previously. It is based on semantic segmentation and does achieve these desired properties. In detail, it is being able to process full VGA images in real-time on a low-power mobile processor. It can further handle multiple image dimensions without retraining, it does not require specific domain knowledge to achieve a high frame rate and it is applicable on a minimal mobile hardware.

We participated in the RoboCup 2018 competition in Montreal with our newly developed BoldBot based on the Darwin-OP and mostly self-printed custom parts. This paper is about the lessons learnt from that competition and further developments for the RoboCup 2019 competition. Firstly, we briefly introduce the team along with an overview of past achievements. We then present a simple, standalone 2D simulator we use for simplifying the entry for new members with making basic RoboCup concepts quickly accessible. We describe our approach for semantic-segmentation for our vision used in the 2018 competition, which replaced the lookup-table (LUT) implementation we had before. We also discuss the extra structural support we plan to add to the printed parts of the BoldBot and our transition to ROS 2 as our new middleware. Lastly, we will present a collection of open-source contributions of our team.

There has always been much motivation for sharing code and solutions among teams in the RoboCup community. Yet the transfer of code between teams was usually complicated due to a huge variety of used frameworks and their differences in processing sensory information. The RoboCup@Home league has tackled this by transitioning to ROS as a common framework. In contrast, other leagues, such as those using humanoid robots, are reluctant to use ROS, as in those leagues real-time processing and low-computational complexity is crucial. However, ROS 2 now offers built-in support for real-time processing and promises to be suitable for embedded systems and multi-robot systems. It also offers the possibility to compose a set of nodes needed to run a robot into a single process. This, as we will show, reduces communication overhead and allows to have one single binary, which is pertinent to competitions such as the 3D-Simulation League. Although ROS 2 has not yet been announced to be production ready, we started the process to develop ROS 2 packages for using it with humanoid robots (real and simulated). This paper presents the developed modules, our contributions to ROS 2 core and RoboCup related packages, and most importantly it provides benchmarks that indicate that ROS 2 is a promising candidate for a common framework used among leagues.

. In this paper we describe the RoboCup Humanoid Kid Size division of team Bold Hearts, the RoboCup team of the University of Hertfordshire, in Hatfield UK. We discuss construction and aspects of our software architecture, as well as some of our solutions to important problems such as action selection, vision, stability, and sensorimotor modeling. Finally, we provide an overview of the relevant experience and achievements of our team.

We study goal-directed behavior in the light of informationally constrained cognition. In a formal information-theoretical model, based on a description of goal-directed behavior as a family of Markov Decision Processes, we study lower bounds of constraints on the information about a goal needed to generate behavior that achieves such a goal at a certain level of optimality. We assume a working memory that operates on this minimally relevant goal information and study the necessary dynamics of in and out flow of information for such a working memory. Finally, we formally analyze explicit constraints on goal information pathways as information bottlenecks. Our results show that intrinsic and behavioral organizations, such as ritualized behavior, salient sub-goals, and natural abstractions, appear as a result of the studied informational constraints. We argue that a closed approach to generate a family of organizational concepts in a coherent way by systematically applying cognitive constraints as taken in this work can constitute an important step toward guiding self-organization.

In this paper we describe the structure and the software of the RoboCup Humanoid Kid Size division team Bold Hearts. We discuss the construction and the aspects of our software architecture, enlightening our solutions to important problems such as the action selection, vision, stability, and sensorimotor modeling. Finally, we provide an overview of the relevant experience and achievements of our team.

This paper deals with the problem of energy constrained gait optimization for bipedal walking. We present a solution to this problem obtained by applying a recently introduced heuristic method, the Alliance Algorithm (AA), and compare its performance against a Genetic Algorithm (GA). We show experimentally that the intrinsic ability of the AA to handle hard constraints enables it to find solutions significantly better than the GA. Also with the constraint removed the AA show more reliable optimization results. Finally, we show that the final gait obtained through this method outperforms most solutions to this problem presented in previous works, in terms of walking speed.

It has been hypothesized that the evolution of sensors is a pivotal driver for the evolution of organisms, and especially, as a crucial part of the perception-action loop, a driver for cognitive development.The questions of why and how this is the case are important: what are the principles that push the evolution of sensorimotor systems?An interesting aspect of this problem is the co-option of sensors for functions other than those originally driving their development (e.g. the auditive sense of bats being employed as a 'visual' modality).Even more striking is the phenomenon found in nature of sensors being driven to the limits of precision, while starting from much simpler beginnings.While a large potential for diversification and exaptation is visible in the observed phenotypes, gaining a deeper understanding of why and how this can be achieved is a significant problem.In this present paper, we will introduce a formal and generic informationtheoretic model for understanding potential drives of sensor evolution, both in terms of improving sensory ability and in terms of extending and/or shifting sensory function.

We consider the effect of the structure of an agent’s cognitive system (its ‘embrainment’) on the organization of its behavior, most notably from the viewpoint of constraints that this structure may impose on the agent’s information processing capabilities. In doing so, we outline the basis of a unified information-theoretic framework to treat the internal organization of decision making for an agent guided by a set of tasks and by the relevance of available information. Using this framework, we show several examples of how starting from the intrinsic considerations of limiting embrainment uncovers a rich spectrum of structure in environment and tasks, which traditional approaches need to specify a-priori, such as salient transition points, task similarity, and local and global organization of the environment. We argue that this structure, which in our approach is inherently relevant to the behavior of an agent, could supply an important guide for the self-organization of an agent charged with a set of tasks.

Based on the fact that information processing is costly, we study in this paper the trade-off between performance and informational requirements. Most importantly, we are interested in how local decisions can alleviate future cognitive burden, measured by the amount of sensory information an agent processes, without conceding performance. We introduce lookahead information as a novel concept to capture the long-term informational requirements and present an iterative method to determine the value of this quantity. Using an example problem, we show how these long-term considerations enable an agent to predict future effects of its actions on its informational burden, and to shape the course of the world to achieve more informationally parsimonious behaviour.

In reinforcement learning problems, the construction of subgoals has been identified as an important step to speed up learning and to enable skill transfer. For this purpose, one typically extracts states from various saliency properties of an MDP transition graph, most notably bottleneck states. Here we introduce an alternative approach to this problem: assuming a family of MDPs with multiple goals but with a fixed transition graph, we introduce the relevant goal information as the amount of Shannon information that the agent needs to maintain about the current goal at a given state to select the appropriate action. We show that there are distinct transition states in the MDP at which new relevant goal information has to be considered for selecting the next action. We argue that these transition states can be interpreted as subgoals for the current task class, and we use these states to automatically create a hierarchical policy, according to the well-established Options model for hierarchical reinforcement learning.

We extend existing models and methods for the informational treatment of the perception-action loop to the case of goaloriented behaviour and introduce the notion of relevant goal information as the amount of information an agent necessarily has to maintain about its goal. Starting from the hypothesis that organisms use information economically, we study the structure of this information and how goal-information parsimony can guide behaviour. It is shown how these methods lead to a general definition and quantification of sub-goals and how the biologically motivated hypothesis of information parsimony gives rise to the emergence of behavioural properties such as least-commitment and goal-concealing.

Hierarchical structuring of behaviour is prevalent in natural and artificial agents and can be shown to be useful for learning and performing tasks. To progress systematic understanding of these benefits we study the effect of hierarchical architectures on the required information processing capability of an optimally acting agent. We show that an information-theoretical approach provides important insights into why factored and layered behaviour structures are beneficial.

The quality of search engines depends usually on the content of the returned documents rather than on the text used to express this content. So ideally, search techniques should be directed more toward the semantic dependencies underlying documents than toward the texts themselves. The most visible examples in this direction are Latent Semantic Analysis (LSA), and the Hyperspace Analog to Language (HAL). If these techniques are really based on semantic dependencies, as they contend, then they should be applicable across languages. To investigate this contention we used electronic versions of two kinds of material with their translations: a novel, and a popular treatise about cosmology. We used the analogy of fingerprinting as employed in forensics to establish whether individuals are related. Genetic fingerprinting uses enzymes to split the DNA and then compare the resulting band patterns. Likewise, in our research we used queries to split a document into fragments. If a search technique really isolates fragments semantically related to the query, then a document and its translation should have similar band patterns. In this paper we (1) present the fingerprinting technique, (2) introduce the material used, and (3) report results of an evaluation for two semantic indexing techniques.