Photovoltaic (PV) energy grows rapidly and is crucial for the decarbonization of electric systems. However, centralized registries recording the technical characteristics of rooftop PV systems are often missing, making it difficult to monitor this growth accurately. The lack of monitoring could threaten the integration of PV energy into the grid. To avoid this situation, remote sensing of rooftop PV systems using deep learning has emerged as a promising solution. However, existing techniques are not reliable enough to be used by public authorities or transmission system operators (TSOs) to construct up-to-date statistics on the rooftop PV fleet. The lack of reliability comes from deep learning models being sensitive to distribution shifts. This work comprehensively evaluates distribution shifts’ effects on the classification accuracy of deep learning models trained to detect rooftop PV panels on overhead imagery. We construct a benchmark to isolate the sources of distribution shifts and introduce a novel methodology that leverages explainable artificial intelligence (XAI) and decomposition of the input image and model’s decision regarding scales to understand how distribution shifts affect deep learning models. Finally, based on our analysis, we introduce a data augmentation technique designed to improve the robustness of deep learning classifiers under varying acquisition conditions. Our proposed approach outperforms competing methods and can close the gap with more demanding unsupervised domain adaptation methods. We discuss practical recommendations for mapping PV systems using overhead imagery and deep learning models.

Compliant and safe human–robot interaction is an important requirement in lower limb exoskeleton design. Motivated by this need, this paper presents the design of a compatible lower limb exoskeleton with variable stiffness actuation and anthropomorphic joint mechanisms, for walking assistance and gait rehabilitation. A novel variable stiffness actuator (VSA) based on a guide-bar mechanism was designed, to provide force and impedance controllability. By changing the crank length of the mechanism, the stiffness of the actuator is adjusted in a wide range (from 0 to 1301 Nm/rad), at fast speed (about 2582 Nm/rad/s), and with low-energy cost. These features make it possible for online stiffness adjustment during one gait cycle, to change the human–robot coupling behavior and improve the performance of the exoskeleton. An anthropomorphic hip joint mechanism was designed based on a parallelogram linkage and a passive joint compensation approach, which absorbs misalignment and improves kinematic compatibility between the human and the exoskeleton joint. Furthermore, a torque control-based multimode control strategy, which consists of passive mode, active mode, and hybrid mode, was developed for different disease stages. Finally, the torque control performance of the actuator was verified by benchtop test, and experimental validations of the exoskeleton with a human subject were carried out, which demonstrate that compliant human–robot interaction was achieved, and stiffness variation benefits for control performance improvement when the control mode changes.

This paper is an extended version of Bílková et al. ((2023b). Logic, Language, Information, and Computation. WoLLIC 2023, Lecture Notes in Computer Science, vol. 13923, Cham, Springer Nature Switzerland, 101–117.). We discuss two-layered logics formalising reasoning with probabilities and belief functions that combine the Łukasiewicz $[0,1]$-valued logic with Baaz $\triangle$ operator and the Belnap–Dunn logic. We consider two probabilistic logics – $\mathsf {Pr}^{{\mathsf {\unicode {x0141}}}^2}_\triangle$ (introduced by Bílková et al. 2023d. Annals of Pure and Applied Logic, 103338.) and $\mathbf {4}\mathsf {Pr}^{{\mathsf {\unicode {x0141}}}_\triangle }$ (from Bílková et al. 2023b. Logic, Language, Information, and Computation. WoLLIC 2023, Lecture Notes in Computer Science, vol. 13923, Cham, Springer Nature Switzerland, 101–117.) – that present two perspectives on the probabilities in the Belnap–Dunn logic. In $\mathsf {Pr}^{{\mathsf {\unicode {x0141}}}^2}_\triangle$, every event $\phi$ has independent positive and negative measures that denote the likelihoods of $\phi$ and $\neg \phi$, respectively. In $\mathbf {4}\mathsf {Pr}^{{\mathsf {\unicode {x0141}}}_\triangle }$, the measures of the events are treated as partitions of the sample into four exhaustive and mutually exclusive parts corresponding to pure belief, pure disbelief, conflict and uncertainty of an agent in $\phi$. In addition to that, we discuss two logics for the paraconsistent reasoning with belief and plausibility functions from Bílková et al. ((2023d). Annals of Pure and Applied Logic, 103338.) – $\mathsf {Bel}^{{\mathsf {\unicode {x0141}}}^2}_\triangle$ and $\mathsf {Bel}^{\mathsf {N}{\mathsf {\unicode {x0141}}}}$. Both these logics equip events with two measures (positive and negative) with their main difference being that in $\mathsf {Bel}^{{\mathsf {\unicode {x0141}}}^2}_\triangle$, the negative measure of $\phi$ is defined as the belief in $\neg \phi$ while in $\mathsf {Bel}^{\mathsf {N}{\mathsf {\unicode {x0141}}}}$, it is treated independently as the plausibility of $\neg \phi$. We provide a sound and complete Hilbert-style axiomatisation of $\mathbf {4}\mathsf {Pr}^{{\mathsf {\unicode {x0141}}}_\triangle }$ and establish faithful translations between it and $\mathsf {Pr}^{\mathsf {\unicode {x0141}}^2}_\triangle$. We also show that the validity problem in all the logics is $\mathsf {coNP}$-complete.

The newly introduced discipline of Population-Based Structural Health Monitoring (PBSHM) has been developed in order to circumvent the issue of data scarcity in “classical” SHM. PBSHM does this by using data across an entire population, in order to improve diagnostics for a single data-poor structure. The improvement of inferences across populations uses the machine-learning technology of transfer learning. In order that transfer makes matters better, rather than worse, PBSHM assesses the similarity of structures and only transfers if a threshold of similarity is reached. The similarity measures are implemented by embedding structures as models —Irreducible-Element (IE) models— in a graph space. The problem with this approach is that the construction of IE models is subjective and can suffer from author-bias, which may induce dissimilarity where there is none. This paper proposes that IE-models be transformed to a canonical form through reduction rules, in which possible sources of ambiguity have been removed. Furthermore, in order that other variations —outside the control of the modeller— are correctly dealt with, the paper introduces the idea of a reality model, which encodes details of the environment and operation of the structure. Finally, the effects of the canonical form on similarity assessments are investigated via a numerical population study. A final novelty of the paper is in the implementation of a neural-network-based similarity measure, which learns reduction rules from data; the results with the new graph-matching network (GMN) are compared with a previous approach based on the Jaccard index, from pure graph theory.

We show that the principal types of the closed terms of the affine fragment of λ-calculus, with respect to a simple type discipline, are structurally isomorphic to their interpretations, as partial involutions, in a natural Geometry of Interaction model à la Abramsky. This permits to explain in elementary terms the somewhat awkward notion of linear application arising in Geometry of Interaction, simply as the resolution between principal types using an alternate unification algorithm. As a consequence, we provide an answer, for the purely affine fragment, to the open problem raised by Abramsky of characterizing those partial involutions which are denotations of combinatory terms.

We present an axiom system for what we call Prior’s Ideal Language and prove its completeness and pure completeness with respect to general models. With this is done, we explain, with examples, why this system provides a useful setting for exploring Arthur Prior’s work.

We revisit the communication primitive in ambient calculi. Previously, such communication was confined to first-order (FO) mode (e.g., merely names or capabilities of ambients can be sent), local mode (e.g., the communication only occurs inside an ambient), or particular cross-hierarchy mode (e.g., parent-child communication). In this work, we explore further higher-order (HO) communication in ambient calculi. Specifically, such a communication mechanism allows sending a whole piece of a program across the borders of ambients and is the only form of communication that can happen exactly between ambients. Since ambients are basically of HO nature (i.e., those being moved may be ambients themselves), in a sense, it appears more natural to have HO communication than FO communication. We stipulate that communications merely occur between equally positioned ambients in a peer-to-peer fashion (e.g., between sibling ambients). Following this line, we drop the local or other forms of communication that violate this criterion. As the workbench, we work on a variant of Fair Ambients extended with HO communication, FAHO. This variant also strengthens the original version in that entirely real-identity interaction is guaranteed. We study the semantics, bisimulation, and expressiveness of FAHO. Particularly, we provide the operational semantics using a labeled transition system. Over the semantics, we define the bisimulation in line with the standard notion of bisimulation for ambients and prove that the bisimulation equivalence (i.e., bisimilarity) is a congruence. In addition, we demonstrate that bisimilarity coincides with observational congruence (i.e., barbed congruence). Moreover, we show that FAHO can encode a minimal Turing-complete HO calculus and thus is computationally complete.