Rambling Ralf Looks at Budworm Optimization

Walters, C.J. (1975). Rambling Ralf Looks at Budworm Optimization. IIASA Working Paper. IIASA, Laxenburg, Austria: WP-75-095

Preview

Text WP-75-095.pdf Download (257kB) | Preview

Abstract

When work on the budworm started at IIASA last year, considerable emphasis was placed on the sad fact that myopic management has led to explosive outbreak conditions. It seems that the government has been spraying hell out of each outbreak area as it appears, with little thought for large scale spatial consequences.

In a great leap sideways, the IIASA groups managed to formalize this myopic viewpoint with site optimization by dynamic programming; it should come as no surprise that (1) the formal myopic solutions closely resemble actual practice (managers are not that stupid), and (2) for New Brunswick as a whole, the myopic optimization still gives poor results (trotting blindly toward a brick wall is not very different from running at it full speed). There have been some attempts to temper the optimization by inserting different objective functions and multiarea constraints on total cutting and spraying, but the results are not very encouraging. Some totally different approach to the optimization is clearly needed, if the study is to avoid going all the way back to brute force searching methods involving gaming simulations.

My interest in this note is to suggest a possible optimization procedure based on the notion that any optimal solution will involve some closed loop "control law" to specify actions as a function of system state (or output observations extracted from the state).

Actions (login required)

View Item