Reading draft · English, with French and English abstracts. This is a working manuscript, not the final deposited thesis.
A clearer path through the thesis
This version combines a chapter-by-chapter review with revised figures, clearer transitions and closer links between explanations and results. The latest visual pass recomposed eight figures and added one, making action and phase, energy exchange, the 2SEG theorem and the two control loops easier to follow.
Manuscript versions
Version
Date
What changed
Read
V7 · latest
Chapter explanations and figure reading revised; eight figures recomposed and one added in the latest visual pass. Transitions and body–appendix organization refined. 116 pages.
Each dated version link keeps its saved PDF unchanged. The latest-manuscript link always points to the newest available version.
The idea
Let the mechanics lead.
01 · Natural motion
A gait the mechanism admits
A pendulum has its own way of oscillating. Can an articulated robot also have its own way of moving forward? In an ideal model without losses or actuation, internal oscillations interact with environmental constraints to produce locomotion.
Joint angles, joint rates and admissible body velocities repeat, while the body advances. This is a natural locomotion cycle: the movement follows the dynamics, rather than a prescribed choreography.
02 · Energy exchange
Two channels, one coupled motion
Just as energy moves back and forth between a pendulum’s kinetic and potential stores, it can move between the robot’s effective oscillatory channel and its propulsive channel.
Ideal environmental constraints redirect this energy without supplying it. The principle requires cyclic propulsion–oscillator exchange (POE), with no net gain or loss in either channel. Both belong to the internal dynamics; they are a different partition from kinetic and potential energy.
03 · Sustain the motion
Supply energy at the right time
With dissipation, an actuator must replace lost energy. Like pushing a swing, timing matters. A phase-locked loop adjusts actuation frequency, while a slower amplitude loop regulates mean speed.
The simulated robot’s mechanics determines the detailed joint motion; the controller does not play back a prescribed gait.
Principle → Method → Control
What the thesis contributes
Principle
A mechanically defined energy exchange
The oscillatory and propulsive energy stores make POE explicit within the internal dynamics. Balanced exchange is necessary for a natural cycle; in general, an energy balance alone does not establish repetition of the whole internal state.
2SEG · theorem
An exact criterion on an oriented section
Observe the joint crossing a fixed angle in the same direction: a Poincaré section. In the regular two-segment model, zero net POE is equivalent to recovery of the same joint rate. If the body velocities also repeat, the complete internal state repeats; nonzero reconstructed translation makes this a locomotion cycle.
The theorem’s conditions
The trajectory follows the original unforced conservative 2SEG equations in nonsingular admissible coordinates, with positive effective inertia and a finite next transverse crossing of the same oriented section. The energy–exchange identity holds along the trajectory. Comparing both body velocities is a separate requirement; energy balance does not supply it. A local NLM interpretation additionally requires a regular family and an embedded phase sweep.
Method · NLM
Compute a repertoire, not one waveform
An auxiliary oscillation seeds a calculation that restores propulsive coupling and solves the full internal periodicity conditions. Continuation then follows neighbouring cycles as frequency and coordination change. A regular family together with its phases forms a local Natural Locomotion Manifold (NLM); mean speed can identify cycles locally, not drive passive migration between them.
3SEG · boundary
One balance cannot determine several joint variables
With three segments and two joints, scalar POE balance remains necessary but is insufficient. Every internal-state component must satisfy its periodicity condition. Numerical results distinguish in-phase and anti-phase organizations at a shared mechanical design, without turning these families into one global manifold.
Control · numerical
Maintain locomotion through local phase–speed control
A local continuation connects one conservative cycle to maintained motion with the same mechanics. A separate, parameter-distinct forced-response study supports local Resonant Locomotion Manifold (RLM) descriptions. PLL timing and slower amplitude regulation maintain simulated motion: fundamental quadrature is a local resonance marker, not a universal rule or a guarantee of global nonlinear stability.
A reading route
Follow the argument.
Start with the question. Read the abstracts and Chapter 1, Letting mechanics organize locomotion.
Understand the principle. Chapters 2–3 move from natural oscillation to natural locomotion, then develop the energy-exchange principle and the 2SEG theorem.
See how motions are computed. Chapter 4, Computing a repertoire of natural gaits, leads into Chapter 5, Two joints, several natural coordinations.
Add losses and control. Chapters 6–7 cover timed energy input and regulation of mean advance. Chapter 8, Synthesis: accompanying the mechanics, brings the argument together.
Consult the seven appendices as needed for models, numerical protocols and geometric detail. The PDF’s linked contents provides chapter navigation.