Act as a senior research professor and systematic literature-review specialist in Agricultural and Biosystems Engineering, Electric Agricultural Machinery, Battery Engineering, Machine Learning, Deep Learning, and Engineering Design Optimization. ## Research Context We are developing an innovative, scientifically rigorous, and publishable research paper on: **Machine Learning and Deep Learning-Assisted Battery Capacity and Physical Pack Dimension Optimization for Electric Agricultural Tractors Under Different Working Durations.** Our primary focus is medium-sized agricultural tractors rated approximately 60–100 kW, operating for 2, 4, 6, 8, and 10 hours under real agricultural working conditions. We aim to determine: 1. Required battery energy capacity (kWh) for different working durations. 2. Battery physical dimensions (length × width × height). 3. Battery-pack volume (L or m³) and mass (kg). 4. Appropriate battery-cell chemistry, capacity, and series/parallel configuration. 5. Realistic cell/module arrangements, clearances, and packaging constraints. 6. Battery feasibility within the physical space available in an agricultural tractor. 7. Maximum achievable working duration for a specified battery installation envelope. 8. Effects of agricultural workload uncertainty on battery sizing. 9. Potential benefits of ML/DL for improving sizing accuracy, robustness, and computational efficiency. Our study must use publicly available experimental datasets. We will not collect our own experimental data. Our computational resources are an Intel Core i7-9750H, 16 GB RAM, and NVIDIA GTX 1660 Ti with 6 GB VRAM. We prefer a physics-guided ML/DL and engineering-optimization approach. FEM, CFD, and GNNs should not be included unless compelling scientific evidence establishes their necessity. ## Task 1 — Comprehensive Literature Discovery Search peer-reviewed research papers, conference papers, review articles, and clearly identified preprints published through October 2026. Prioritize the following research themes: - Battery sizing and selection for electric agricultural tractors. - Battery energy requirements for different field-working durations. - Agricultural tractor load cycles and energy-demand prediction. - Physical battery-pack sizing, geometric layout optimization, and module packaging. - Battery capacity–mass–volume trade-offs. - Tractor installation-space constraints and electrification feasibility. - ML/DL-based tractor power and energy-demand forecasting. - Probabilistic and uncertainty-aware battery sizing. - Physics-informed ML/DL for battery or vehicle energy-system design. - Battery design in non-road mobile machinery, including construction and mining equipment. Identify at least 20–30 highly relevant papers where available, prioritizing the strongest 10–15 for detailed comparison. Do not invent publications or DOIs. Verify bibliographic information and distinguish peer-reviewed papers from preprints. ## Task 2 — Detailed Comparative Evidence Matrix For each closely related study, extract: 1. Authors, publication year, paper title, journal/conference, and DOI. 2. Agricultural machine or vehicle category. 3. Tractor rated power (kW), where reported. 4. Agricultural operation and working duration. 5. Public, private, experimental, or simulated dataset used. 6. Battery-sizing methodology and governing equations. 7. Battery capacity, mass, volume, and physical dimensions reported. 8. Whether actual cell/module arrangements were optimized. 9. Whether realistic installation-space constraints were considered. 10. ML/DL algorithms and their role, if any. 11. Whether uncertainty and reliability were modeled. 12. Optimization objectives and constraints. 13. Experimental validation methods and performance metrics. 14. Principal findings and numerical results. 15. Important limitations and unresolved engineering questions. Provide a structured comparison table and citations for all extracted findings. Use "Not reported" or "Not verified" when the source does not support a claim. Do not infer that a method was absent merely because it is not mentioned in the abstract. ## Task 3 — Scientific Research-Gap Identification Identify gaps specifically related to integrating: **Agricultural field-load variability + ML/DL prediction + work-duration-dependent battery sizing + actual three-dimensional battery-pack geometry + tractor installation feasibility.** Evaluate whether previous research has already developed this complete combination. Distinguish between: - Established engineering methods. - Incremental improvements. - Potentially original methodological contributions. - Missing validation evidence. - Genuine, testable scientific gaps. Investigate whether researchers have developed bidirectional models: A. Given required working hours, predict the minimum feasible battery capacity and physical dimensions. B. Given limited installation space and battery mass constraints, predict the maximum achievable tractor working duration. Examine whether similar approaches exist in automotive, construction, mining, and other electrified off-highway machinery. Do not claim that a gap is novel simply because a single paper does not address it. ## Task 4 — ML/DL Methodology Assessment Critically evaluate the scientific suitability of: - Random Forest and XGBoost. - GRU and LSTM. - Temporal Convolutional Networks (TCN). - Lightweight Transformer models. - Probabilistic deep learning and quantile regression. - Physics-guided or physics-informed machine learning. - Hybrid ML/DL + constrained engineering optimization. Determine which approach is most appropriate for predicting uncertain agricultural energy and power requirements from real tractor measurements. Investigate how models can be evaluated on unseen agricultural fields and tractors. Discuss the risks of small independent datasets, time-series leakage, unrealistic simulation-generated labels, poor uncertainty calibration, and domain shifts between diesel and battery-electric tractors. Do not recommend deep learning solely because it is more sophisticated than conventional machine learning. ## Task 5 — Public Experimental Datasets Find and verify downloadable public datasets suitable for the proposed study. Start with: - TUM Agricultural Load Cycles: https://zenodo.org/records/14619787 - Tümosan 81.110 agricultural tractor dataset: https://data.mendeley.com/datasets/bbjtjrsxvx/3 - CALB prismatic lithium-ion battery dataset: https://data.mendeley.com/datasets/ycx459r5c3/2 Identify additional relevant agricultural tractor load datasets, electric tractor experiments, battery-cell specification datasets, and publicly available tractor packaging or CAD dimensions. For each dataset report: - Official download URL and DOI. - Experimental origin and measured variables. - Number of independent recordings. - Tractor type and rated power. - Available working durations. - Sampling frequency and file formats. - Licensing and accessibility. - Suitability for ML/DL training. - Suitability for independent testing. - Data limitations. Prioritize genuinely measured agricultural operating data over synthetic datasets. ## Task 6 — Battery Geometry and Installation Feasibility Investigate established methods for calculating and optimizing: - Number of cells in series and parallel. - Battery-cell and module dimensions. - Pack length, width, and height. - Electrical connection and cell-spacing requirements. - Cooling and enclosure volume allowances. - Pack mass and volume. - Geometric packing efficiency. - Tractor installation-envelope restrictions. Find studies that move beyond approximate gravimetric or volumetric energy density to physically realizable module layouts. Identify publicly available tractor installation-space measurements or dimensioned engineering drawings, especially for 60–100 kW tractors. Distinguish a conceptual battery layout from a manufacturer-validated, mechanically installable pack. ## Task 7 — Generate and Rank Research Ideas Develop 5–7 original, scientifically defensible research-paper candidates. For each candidate provide: 1. Proposed publication-quality title. 2. Specific research gap and evidence supporting it. 3. Research question and testable hypotheses. 4. Potential scientific novelty. 5. Public experimental datasets needed. 6. ML/DL methods and appropriate baselines. 7. Physical battery-sizing and geometric-optimization methodology. 8. Required equations and optimization constraints. 9. Experimental validation strategy. 10. Potential journal suitability. 11. Main scientific and practical risks. 12. Feasibility on our available computer hardware. Rank the candidates using a transparent evaluation rubric covering scientific novelty, data accessibility, validation quality, engineering relevance, computational feasibility, and methodological complexity. Clearly identify the strongest candidate and explain why. ## Task 8 — Critical Recommendation Recommend whether our strongest paper should prioritize: A. Probabilistic ML/DL for agricultural battery-energy requirements. B. ML/DL-assisted physical battery-pack dimension and capacity optimization. C. Integrated bidirectional battery sizing: required working hours ↔ maximum physically achievable working duration. D. Another more scientifically original alternative uncovered by the literature. Be critical. If any proposed contribution is already well established, identify the competing papers and propose a more defensible alternative. ## Expected Output Produce a comprehensive, scholarly research-gap report containing: - Verified literature comparison matrix. - Closest competing studies. - Research-gap evidence map. - Public dataset suitability matrix. - ML/DL methodology recommendations. - Battery-pack geometry and packaging considerations. - Five to seven ranked publication candidates. - Strongest recommended research concept. - A novelty-risk assessment. - Explicit unresolved questions to investigate before finalizing the research title. Cite all factual findings with verifiable papers, DOIs, and source links. Separate source-supported conclusions from proposed ideas and hypotheses. **Important: We are currently brainstorming and evaluating originality. Do not produce a complete research proposal, implementation code, or software-installation guide yet. Our goal is to select the strongest scientifically publishable research idea first.**