Publications

Journal of Advanced Research in Applied Sciences and Engineering Technology

Vol. 57, Issue 4 (2026)

Non-Classical Optimal Control with Piecewise Royalty Functions

This paper presents a hybrid Newton-Brent shooting method for non-classical optimal control problems with unknown final state and non-differentiable piecewise royalty structures. We introduce hyperbolic tangent approximations to replace discontinuous functions, derive natural boundary conditions from Malinowska & Torres (2010), and validate against four discretization schemes. The C++ implementation with AMPL/MINOS cross-verification demonstrates superior accuracy for three-stage royalty structures.

Key Technical Contributions

• Tanh smoothing for piecewise discontinuities with controllable approximation parameter
• Novel shooting method combining Newton iteration with Brent's method for boundary conditions
• Four-scheme validation: Euler, Runge-Kutta, Trapezoidal, Hermite-Simpson

Real-World Applications

Licensing agreements, resource extraction royalties, technology monetization with threshold-based payment structures.

Journal of Advanced Research in Applied Sciences and Engineering Technology

Vol. 59, Issue 1 (2026)

Four-Stage Royalty Optimization via Modified Shooting

This work extends our three-stage framework to four-stage piecewise systems using a Newton-Golden Section hybrid shooting method. We formalize the Hamiltonian system for non-differentiable integrands, derive stationarity conditions and costate dynamics under tanh smoothing, and demonstrate convergence across smoothing parameters (k=50, k=250). Final costate validation via integral boundary conditions confirms mathematical rigor.

Key Technical Contributions

• Extension of hybrid shooting to four-stage piecewise objective functions
• Formal derivation of costate dynamics for smoothed non-differentiable systems
• Golden Section Search integration for improved convergence in boundary value problems

Impact

Demonstrates that structured numerical methods outperform black-box solvers for problems with strategic threshold behavior common in business and policy applications.

Multidisciplinary Science Journal

e2026038 (2026)

National Rice Production Forecasting for Malaysia

This study evaluates four forecasting methodologies for Malaysian national rice production: Naïve, Holt-Winters (HLT), ARIMA, and Feedforward Neural Networks (FBNN). Using training data from 1961-2007 and validation from 2008-2019, we assess accuracy via MAE, RMSE, MAPE, and Theil's Inequality Coefficient. Full residual diagnostics reveal that HLT achieves optimal validation performance with mild overestimation bias—preferred for buffer-stock planning in net-importing economies.

Key Technical Contributions

• Comprehensive four-method comparison with explicit model selection criteria
• Policy-aligned accuracy metrics favoring conservative (overestimating) forecasts
• Full residual diagnostics ensuring statistical validity

Policy Significance

Provides evidence-based framework for strategic import scheduling in net-importing ASEAN economies. Cited in ASEAN food security working groups.