The article deals with the concept of building an automated system for the harvesting of apple crops. This system is a robotic complex mounted on a tractor cart, including an industrial robot and a packaging system with a container for fruit collection. The robot is equipped with a vacuum gripper and a vision system. A generator for power supply, a vacuum pump for the gripper and an equipment control system are also installed on the cart. The developed automated system will have a high degree of reliability that meets the requirements of operation in the field.

An algorithm for parallel calculations in a dynamic model of manipulation robots obtained by the Lagrange–Euler method is developed. Independent components were identified in the structure of the dynamic model by its decomposition. Using the technology of object-oriented programming, classes corresponding to the structures of the selected components of the dynamic model were described. The algorithmization of parallel computing is based on the independence of the calculation of objects of individual classes and the sequence of matrix operations. The estimation of the execution time of parallel algorithms, the resulting acceleration, and the efficiency of using processors is given.

Spatial reasoning in multimodal large language models (MLLMs) remains vulnerable to hallucination: a model may return a fluent answer while violating metric, geometric, or cross-task constraints. This failure mode is particularly consequential when video-based spatial estimates inform navigation, manipulation, or human oversight. We present COSMES, an inference-time reliability framework for Spatial-MLLMs that connects six complementary mechanisms in one pipeline: geometry-proxy (depth-aware) frame sampling, semantic–geometric dual-branch aggregation, spatial chain-of-thought verification, geometric self-correction, hallucination-pattern detection, and stochastic consistency-based uncertainty estimation. The central contribution is the integration of input-, reasoning-, and output-level safeguards around a frozen spatial MLLM, rather than treating hallucination as only an object-presence or decoding error. COSMES returns not only a task answer but also inspectable reliability evidence, including correction records, consistency warnings, sampling dispersion, and uncertainty intervals. These outputs can provide an operational basis for abstention, re-observation, or human verification when paired with a validated downstream policy. The present evaluation is offline and does not establish calibrated confidence, broad generalization, real-time performance, or safety in deployment.

A method for compiling object schemes is proposed, which allows constructing algorithms for calculating the kinematic parameters of robotic manipulators. Examples of compiling object schemes for calculating the velocities and accelerations of points selected on the links of the robotic manipulator are considered. An analysis of the computational complexity of the obtained algorithms is carried out and a method for increasing their computational efficiency is proposed. An increase in computational efficiency is achieved based on the use of the associativity property due to the reduction of additional and multiplication operations performed by the algorithm. Graphs of computational processes illustrating the developed algorithms are presented. The developed algorithms allow parallel calculations; this will further increase the efficiency of calculations when using multiprocessor computing systems. As a result of the study, based on the object approach, an effective universal method for calculating the kinematic parameters of robotic manipulators has been developed. This will improve the quality of robot control.

To solve the limitations of single optimization algorithms, such as premature convergence, insufficient global exploration, and high susceptibility to local optima, a Hybrid Optimization Algorithm (HOA) based on multi-subpopulation collaboration and multi-strategy fusion is proposed. The HOA uses Logistic chaotic mapping for population initialization to enhance uniformity and diversity. The population is then divided into four subpopulations; each is optimized independently using different strategies, including the genetic algorithm (GA), Gray Wolf Optimizer (GWO), self-attention mechanism, and k-nearest neighbor graph (kNN). This design leverages the strengths of individual algorithms while mitigating their respective limitations. An elite information exchange mechanism facilitates knowledge transfer by randomly reassigning elite individuals across subpopulations at fixed iteration intervals. Additionally, global optimization strategies including differential evolution (DE), Simulated Annealing (SA), Local Search (LS), and time of arrival (TOA) position adjustment are integrated to balance exploration and exploitation, thereby enhancing convergence accuracy and the ability to escape local optima. Evaluated on the CEC2017 benchmark suite and real-world engineering problems, the HOA demonstrates superior performance in convergence speed, accuracy, and robustness compared to single-algorithm approaches—notably, HOA ranks 1st in 30-dimensional CEC2017 functions. By effectively integrating multiple optimization strategies, the HOA provides an effective and reliable solution to complex optimization challenges.