Electric vehicles (EVs) are characterized by a significant variety of possible powertrain configurations, ranging from one to four electric machines, which can have an on-board or in-wheel layout. Multiple models of production EVs have recently been introduced on the market, with 4-wheel-drive (4WD) architectures based on a central motor within each axle, connected to the wheels through a gearbox, a differential, and half-shafts. In parallel, an important body of research and industrial demonstrations have covered the topic of 2-speed transmission systems for EVs, with the target of enhancing longitudinal acceleration and gradeability performance, while increasing the operating efficiency of the electric powertrain. Although several recent studies compare different electric powertrain architectures, to the best of the authors’ knowledge the literature misses a comparison between 2-wheel-drive (2WD) and 4WD configurations for the same EV, from the viewpoint of drivability and energy consumption. This paper targets this gap, by assessing 2WD and 4WD powertrain layouts with central motors, for a case study light passenger car for urban mobility, including consideration of the effect of single- and 2-speed transmission systems. An optimization routine is used to calculate the energy-efficient gear state and/or torque distribution for each considered configuration. For the specific EV, the results highlight the favourable trade-off of the single-speed 4WD layout, capable of reducing the energy consumption during driving cycles by approximately 9% with respect to the conventional 2WD layout with single-speed transmission, while providing satisfactory drivability and good gradeability, especially in low tire–road friction conditions.

State-of-the-art antilock braking systems (ABS) are reactive, i.e., they activate after detecting that wheels tend to lock in braking. With vehicle-to-everything (V2X) connectivity becoming a reality, it will be possible to gather information on the tire–road friction conditions ahead, and use these data to enhance wheel slip control performance, especially during abrupt friction level variations. This study presents a nonlinear model predictive controller (NMPC) for ABS with preview of the tire–road friction profile. The potential benefits, optimal prediction horizon, and robustness of the preview algorithm are evaluated for different dynamic characteristics of the brake actuation system, through an experimentally validated simulation model. Proof-of-concept experiments with an electric vehicle prototype highlight the real-time capability of the proposed NMPC ABS, and the associated wheel slip control performance improvements in braking maneuvers with high-to-low friction transitions.

Physical AI requires machines to sense, decide and act under tight constraints of energy, latency, safety and robustness. A fly escaping an approaching hand captures the core principle: a fast sensor-action reflex acts before full deliberation, while higher neural resources remain available for richer behavior. This Perspective proposes a layered Reflex-Policy architecture in which reflex layers execute fast, local, ADC-light actions near sensors and actuators, while policy layers perform slower learning, planning, optimization and rule updates. The two are mutually protective collaborators: policy layers define safe operating envelopes, while reflex layers shield policy processors from high-frequency events and avoidable data floods. We position binary spintronic MTJ crossbars as a plausible technology path for low reflex layers, while distinguishing today's MRAM/eMRAM technologies from the proposed reflex-layer crossbar module, which remains an architectural research target. The contribution is a system-integration framework: intelligence is distributed along the sensing, energy, storage and action chain, and each event is assigned to the lowest sufficient layer. We formalize Energy Returned on Invested Energy for Embodied Intelligence (EROIE), compare the framework with related architectures, and state its limitations and validation needs.

A broad literature deals with torque-vectoring (TV) algorithms based on direct yaw moment control (DYC), which continuously generate longitudinal tire forces that are different on the left and right vehicle sides. Such set-ups enable desirable yaw rate and sideslip responses in normal and extreme driving conditions. However, they imply two electric powertrains per axle, or controllable differentials, and therefore the related controllers are not applicable to typical production electric vehicle layouts with a single centralized electric machine and open differential per axle. Nevertheless, in proximity of the limit of handling, the cornering response of conventional all-wheel-drive electric vehicles with one motor per axle can benefit from front-to-rear axle torque-vectoring (ATV), which, by controlling the longitudinal force distribution between the front and rear axles, indirectly modifies the yaw moment contribution due to the lateral tire forces. Only a few studies exploit this option, with limited coverage, e.g., without considering the energy efficiency aspects and the electronic stability control (ESC) function actuating the friction brakes. To address the gap, this paper presents real-time capable nonlinear model predictive controllers (NMPCs) for ATV and ESC, targeting: i) energy consumption reduction in normal driving, through powertrain and tire slip power loss limitation; and ii) yaw rate, sideslip angle and wheel slip control at the limit of handling, by considering the trade-off between powertrain and friction brake actuation. The performance is assessed through simulations with a validated vehicle model, and experiments on a vehicle demonstrator. With respect to benchmarking rule-based controllers, the results show that the NMPCs bring significant reductions of the yaw rate errors and friction brake power losses, while increasing the exit speed in acceleration-during-cornering conditions.

Physical AI systems must respond to real-world events under strict constraints of time, energy and safety. This Perspective clarifies the distinct roles of latency, throughput, bandwidth and world models within a Reflex–Policy architecture and argues that increasingly powerful policy and world models do not eliminate the need for well-designed local reflex layers. On the contrary, reflex layers can dramatically reduce upstream data rates, preserve hard real-time safety, and allow policy models to focus on prediction, planning and rare events. A binary spintronic reflex crossbar has been demonstrated in proof-of-concept form for fast shadow bypass in photovoltaic systems and for battery cell balancing, confirming that non-volatile magnetic rule fabrics can perform sub-millisecond, parallel switching decisions in real physical domains. Multilevel spintronic devices align better with weighted inference on the policy side. Quantitative examples from robotics, automotive and energy systems illustrate how this layered approach can reduce communication bandwidth by 1–3 orders of magnitude while maintaining safety and efficiency. A comparison with existing real-time control technologies, safety microcontrollers, FPGAs and neuromorphic processors, clarifies where spintronic reflexes offer distinct advantages in standby power, non-volatility and inspectable rule logic.

Insects achieve millisecond sensor–motor loops with tiny sensors, compact neural circuits, and powerful actuators, embodying the principles of Edge AI. We present a comprehensive architectural blueprint translating insect neurobiology into a hardware–software stack: a latency-first control hierarchy that partitions tasks between a fast, dedicated Reflex Tier and a slower, robust Policy Tier, with explicit WCET envelopes and freedom-from-interference boundaries. This architecture is realized through a neuromorphic Reflex Island utilizing spintronic primitives, specifically MRAM synapses (for non-volatile, innate memory) and spin-torque nano-oscillator (STNO) reservoirs (for temporal processing), to enable instant-on, memory-centric reflexes. Furthermore, we formalize the biological governance mechanisms, demonstrating that, unlike conventional ICEs and miniturbines that exhibit narrow best-efficiency islands, insects utilize active thermoregulation and DGC (Discontinuous Gas Exchange) to maintain nearly constant energy efficiency across a broad operational load by actively managing their thermal set-point, which we map into thermal-debt and burst-budget controllers. We instantiate this integrated bio-inspired model in an insect-like IFEVS thruster, a solar cargo e-bike with a neuromorphic safety shell, and other safety-critical edge systems, providing concrete efficiency comparisons, latency, energy budgets, and safety-case hooks that support certification and adoption across autonomous domains.